WO2011053790A2

WO2011053790A2 - Assay of closely linked targets in fetal diagnosis and coincidence detection assay for genetic analysis

Info

Publication number: WO2011053790A2
Application number: PCT/US2010/054728
Authority: WO
Inventors: Bernhard Georg Zimmermann; Marc Unger
Original assignee: Fluidigm Corporation
Priority date: 2009-10-30
Filing date: 2010-10-29
Publication date: 2011-05-05
Also published as: WO2011053790A3

Abstract

Methods for detecting chromosomal aneuploidy of a specified chromosome or chromosome region are provided. Also provided are methods for genetic analysis of heterogeneously sized chromosomal DNA fragments. The methods are useful for non-invasive prenatal diagnosis and other genetic analyses.

Description

ASSAY OF CLOSELY LINKED TARGETS IN FETAL DIAGNOSIS AND COINCIDENCE DETECTION ASSAY FOR GENETIC ANALYSIS

CROSS-REFERENCES TO RELATED APPLICATIONS

[0001 ] The present application claims priority to U.S. Provisional Application No. 61/256,514, filed October 30, 2009, entitled Coincidence Detection Assay for

Genetic Analysis, and U.S. Provisional Application No. 61/256,522, filed October 30, 2009, entitled Assay of Closely Related Targets in Fetal Diagnosis, each of which is herein incorporated by reference in their entirety for all purposes.

FIELD OF THE INVENTION

[0002] The present invention relates generally to the area of assays for the analysis of fetal DNA. In particular, the present invention relates to assays of closely linked targets in fetal diagnosis and coincidence detection assays for genetic analysis.

BACKGROUND

[0003] It is well established that maternal peripheral blood contains circulating cell- free DNA (cf-DNA) derived from both maternal cells (maternal cf-DNA) and fetal/placenta cells (fetal cf-DNA). Analysis of circulating cf-DNA of fetal/placental origin has great potential as a diagnostic tool to detect fetal genetic traits, including aneuploidy and other genetic abnormalities.

[0004] However, there are barriers to reliable analysis of fetal cf-DNA. One is that fetal DNA represents only about 3-10% of total cell-free DNA in maternal circulation. Thus, analysis of fetal sequences (other than Y-chromosome sequences) requires a method to distinguish fetal DNA from a large excess of maternal DNA. In the case of diagnosis of aneuploidy, there is the additional challenge of determining the dose of a chromosomal sequence in cell-free material.

[0005] A number of approaches have been proposed for analysis of fetal cf-DNA. For example, one strategy involves enrichment of fetal sequences by size fractionation to reduce the maternal background. Li et al., 2004, Clinical Chemistry 50:1002-1 1 , reported that the most fetal circulating cell-free DNA molecules have a molecular size less than 300 bases, while the majority of maternally derived sequences are larger than fetal derived sequences, often larger than 1 kb, with a substantial portion greater than 10 kb in length (e.g., in the 10-23 kb size range).

[0006] In another approach, "digital PCR" has been proposed for molecular detection of fetal chromosomal aneuploidy. Lo et al., 2007, PNAS 104:131 16-121 . Using probes for chromosomes 1 and 21 , Lo et al. measured the copy number of chromosome 21 relative to chromosome 1 (a reference chromosome) to identify samples in which chromosome 21 was over represented. Using this method, Lo et al. calculated that, provided with an enriched DNA sample in which 25% of the DNA was fetal cf-DNA, 7,600 digital PCR reactions would allow correct ploidy classification 97% of the time. Quake et al. also described use of digital PCR to detect chromosome 21 trisomy, see PCT Pat. Pub. WO 2007/092473.

[0007] Notwithstanding the considerable potential of cf-DNA analysis for fetal genetic diagnosis, there is a need for more sensitive, reliable and economical methods for analysis of fetal DNA, particularly detection of aneuploidy. The present invention provides this and other benefits.

BRIEF SUMMARY

[0008] In one aspect the invention provides a method for detecting chromosomal aneuploidy of a specified chromosome or chromosome region by (a) providing a sample comprising cell-free DNA (cf-DNA) obtained from a biological fluid of a subject, (b) partitioning the sample into a plurality of reaction volumes and assaying each reaction volume for the presence of each of two or more DNA target sequences, wherein each of said two or more DNA target sequences is closely linked to at least one other of said two or more DNA target sequences in the specified chromosome or chromosome region, wherein the sample is partitioned so that most reaction volumes comprise not more than 10 discrete DNA fragments comprising a target sequence, (c) determining the number of reaction volumes that contain at least one DNA molecule comprising at least one of said two or more of DNA target sequences, (d) comparing said number of reaction volumes with a reference number characteristic of a known or assumed chromosome ploidy, wherein when said number of reaction volumes is higher or lower than the reference number it is indicative of chromosomal aneuploidy of the specified chromosome or chromosome region. In one embodiment the biological fluid is plasma.

[0009] In an exemplary embodiment the subject is a pregnant woman and the presence or absence of a chromosomal aneuploidy of a specified chromosome or chromosome region of the fetus is determined. In one embodiment the invention provides a method for diagnosing fetal chromosomal aneuploidy of a specified chromosome or chromosome region by (a) providing a sample comprising circulating cell-free DNA (cf-DNA) obtained from human maternal plasma, (b) partitioning the sample into a plurality of reaction volumes, (c) assaying each reaction volume for the presence of each of two or more DNA target sequences, wherein said target sequences are closely linked in the specified chromosome or chromosome region, (d) determining the number of reaction volumes that contain at least one DNA molecule comprising at least one of said two or more of DNA target sequences, and e) comparing said number of reaction volumes with a reference number characteristic of a known or assumed chromosome ploidy, wherein when said number of reaction volume is higher or lower than the reference number it is indicative of fetal chromosomal aneuploidy of the specified chromosome or chromosome region. In some embodiments, the closely linked target sequences are on chromosome 18 or 21 .

[0010] In an exemplary embodiment the subject is an individual diagnosed with or suspected of having cancer. In some embodiments the closely linked target sequences are in the ATM (ataxia telangiectasia mutated) locus.

[001 1 ] The sample may be partitioned so that most reaction volumes comprise not more than 5 discrete DNA fragments comprising a target sequence. In some embodiments about 10% and about 90% of the reaction volumes assayed provide a positive signal indicating the presence of at least one target sequence in the reaction volume. In some embodiments on average each reaction volume contains about 0.2 to about 10 DNA fragments comprising one or more target sequences. In some cases, on average each reaction volume contains cf-DNA in an amount equal to that found in about 0.02 to about 10 microliters plasma.

[0012] The reference number characteristic of a known chromosome ploidy may be determined by a simultaneous assay of multiple target sequences on a reference chromosome that is other then the specified chromosome, or on a chromosome region that is other then the specified chromosome region. [0013] In some embodiments the sample comprising cf-DNA is enriched for DNAs smaller than 500 bp.

[0014] In some embodiments, at least 5 closely linked target sequences, sometimes at least 10 closely linked target sequences are assayed.

[0015] In some embodiments each target sequence is less than 300 bases in length, sometimes less than 100 bases in length.

[0016] In some embodiments the closely linked target sequences all are within 2 kb of each other on the chromosome. In some cases at least 10 target sequences are assayed and each target sequence is closely linked to at least one of the other target sequences. For example, each target sequence may be located in the chromosome within 500 bp of at least one of the other target sequences.

[0017] In one aspect, the invention provides a method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

a) distributing a DNA sample comprising fragmented chromosomal DNA heterogeneous in size into a plurality of reaction volumes, wherein after said distribution on average each reaction volume contains less than one DNA fragment containing a first specified chromosomal target sequence,

wherein said sample comprises a first population of DNA fragments in a first size range and a second population of DNA fragments in a second size range, and the mean of the second size range is greater than the mean of the first size range; b) assaying each reaction volume for the presence of at least one fragment containing said first specified chromosomal target sequence, wherein if such a fragment is present a first detectable signal is produced, and assaying each reaction volume for the presence of a fragment containing a second specified chromosomal target sequence, wherein if such a fragment is present a second detectable signal is produced,

wherein said first and second specified chromosomal target sequences are closely linked on the same chromosome and wherein said first and second detectable signals are distinguishable;

c) determining from the number or proportion of reaction volumes in which the first detectable signal is produced and no second detectable signal is produced the presence or representation of fragments in the DNA sample that (i) comprise the first specified chromosomal target sequence and (ii) have a length in the first size range, and/or

determining from the number or proportion of reaction volumes in which both the first and second detectable signals are produced with the presence or representation number of fragments in the DNA sample that (i) comprise the first specified chromosomal target sequence and (ii) have a length in the second size range.

[0018] In a related embodiment the invention provides a method of differentially detecting one or more target sequences in a sample comprising a mixture of short and long polynucleotides, the method comprising:

distributing the sample into discrete reaction volumes, each volume comprising an average of no more than about one detectable polynucleotide per reaction volume;

distributing in each reaction volume at least a first detectable probe and a second detectable probe wherein the first and second probes are selected such that both the first and second probes will bind to a long polynucleotide and only the first or the second probe will bind to a short polynucleotide;

detecting first and/or second probe binding to polynucleotide in a reaction volume, wherein detection of either the first or second probe binding indicates detection of probe binding to a short polynucleotide and detection of both first and second probe binding indicates detection of probe binding to a long polynucleotide, wherein a "detectable polynucleotide" is a polynucleotide that can be detected with the first probe and/or the second probe.

[0019] In this embodiment the probes may detect amplicons generated in the reaction volumes by amplification of sequences from said short and long polynucleotides.

[0020] In one embodiment the invention provides a method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

a) distributing a DNA sample comprising fragmented chromosomal DNA heterogeneous in size into a plurality of reaction volumes, wherein after said distribution on average each reaction volume contains an amount of DNA equal to less than 1 haploid genome equivalent of DNA,

wherein said sample comprises a first population of DNA fragments in a first size range and a second population of DNA fragments in a second size range, and the mean of the second size range is greater than the mean of the first size range;

b) assaying each reaction volume for the presence of one or more target sequences from a set of first target sequences, wherein if a fragment comprising one or more of said first target sequences is present in the reaction volume a signal A is detected;

c) assaying each reaction volume for the presence of one or more target sequences from a set of second target sequences, wherein if a fragment comprising one or more of said second target sequences is present in the reaction volume a signal B is detected;

wherein each first target sequence is closely linked to at least one second target sequence on a chromosome;

wherein in a formula describing a relationship of assays

"-A-" indicates that a target sequence in the first set is assayed;

"-B-" indicates that a target sequence in the second set is assayed; and the order of symbols in a formula corresponds to the order of said target sequences on said chromosome,

each reaction volume is assayed for target sequences having the relationship (-A-B-)N, wherein N = 1 -10 and each target sequence is independently selected; distinguishing reaction volumes in which a single signal is detected from reaction volumes in which more than one signal is detected, wherein a reaction volume in which a single signal is detected is counted as containing a fragment that comprises a target sequence and is from said first population and a reaction volume in which both signals are detected is counted as containing a fragment that comprises a target sequence and is from said second population.

[0021 ] In an embodiment the method includes optionally assaying each reaction volume for the presence of one or more of a set of third target sequences, wherein if a fragment comprising one or more of said third target sequences is present in the reaction volume a signal C is detected;

wherein each third target sequence is closely linked to at least one first target sequence and/or at least one second target sequence on said chromosome;

"-C-" indicates that a target sequence in the third set is assayed;

the order of symbols in a formula corresponds to the order of said target sequences on said first chromosome, each reaction volume is assayed for target sequences having one or more relationships selected from:

i) (-A-B-)N,

ii) C-(-A-B-)_N-C

iii) A-(-B-)_N-A

iv) (-A-B-C-)N

v) (-A-B-B-)_N

vi) combinations of (i) - (v) wherein each target sequence in the combination is closely linked to at least one other target sequence in the combination vii) Two or more groups of assays selected from (i) - (v), wherein each group is independently selected and target sequences of one group are not necessarily closely linked to a target sequence in another group,

wherein N = 1 -10 and each target sequence is independently selected;

distinguishing reaction volumes in which a single signal is detected from reaction volumes in which more than one signal is detected, wherein a reaction volume in which a single signal is detected is counted as containing a fragment that comprises a target sequence and is from said first population and a reaction volume in more than one signal is detected is counted as containing a fragment that comprises a target sequence and is from said second population.

[0022] In one embodiment the invention provides a method for genetic analysis comprising:

a) carrying out the method of claim 4 in which the target sequences are on a first chromosome

b) carrying out the method of claim 13 in which the target sequences are on a second chromosome, with the modification in assaying target sequences on the second chromosome signal A or signal B or both is distinguishable from signal A or signal B or both, respectively, on the first chromosome

comparing the representation of fragments from the first and/or second populations in (a) and representation of fragments from the first and/or second populations in (b).

[0023] In one embodiment the invention provides a method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

wherein said DNA sample comprises a first population of DNA fragments in a first size range and a second population of DNA fragments in a second size range, and the mean of the second size range is greater than the mean of the first size range;

b) assaying each reaction volume for the presence of a first target sequence, wherein if a fragment comprising one or more of said first target sequences is present in the reaction volume a signal A is detected;

c) assaying each reaction volume for the presence of a second target sequence, wherein if a fragment comprising one or more of said second target sequences is present in the reaction volume a signal A' (A prime) is detected

wherein the first and second target sequences are allelic variants in the same gene

d) assaying each reaction volume for the presence of one or more of a set of third target sequences, wherein if a fragment comprising one or more of said third target sequences is present in the reaction volume a signal B is detected;

e) optionally assaying each reaction volume for the presence of one or more of a set of forth target sequences, wherein if a fragment comprising one or more of said fourth target sequences is present in the reaction volume a signal C is detected;

wherein each first target sequence is closely linked to at least one third target sequence, each second target sequence is closely linked to at least one third target sequence and/or at least one fourth target sequence;

wherein in a formula describing a relationship of assays

"-A-" indicates that a first target sequence is assayed;

"-A -" indicates that a second target sequence is assayed;

"-B-" indicates that a third target sequence is assayed;

"-C- " indicates that a fourth target sequence is assayed; and

the order of symbols in a formula corresponds to the order of said target sequences on said first chromosome,

the assays in each reaction volume comprises one or more relationships selected from:

i) B-A, ii) A-B,

iii) B-A-C,

iv) B-A-B,

and one or more relationships selected from:

v) B-A\

vi) A'-B;

vii) B-A'-C

viii) B-A'-B

ix) C-A'-C

identifying reaction volumes in which only a signal selected from A is detected as comprising a fragment from the first population, said fragment comprising the first target sequence;

identifying reaction volumes in which only a signal selected from A' is detected as comprising a fragment from the first population, said fragment comprising the second target sequence.

[0024] In various embodiments target sequences may be considered closely linked on the chromosome if they are separated by less than 1 kbp. In various embodiments the DNA sample is circulating cell-free DNA from a human, such as a pregnant woman or person diagnosed with or suspected of having, cancer. In certain embodiments DNA fragments in the first size range are enriched in fetal cf- DNA compared to total cf-DNA and DNA fragments in a second size range are enriched in maternal cf-DNA compared to total cf-DNA. In some embodiments the first size range is 25-200 bp and the second size range is 400 bp to 20 kbp. In various embodiments the first and second detectable signals are fluorescence, and may be a measure of melting temperature. In some embodiments the combined lengths of the first specified chromosomal sequence, the second specified chromosomal sequence, and the distance between said sequences is more than 100 bp and less than 1000 bp.

BRIEF DESCRIPTION OF THE FIGURES

[0025] Figures 1A and 1 B show a schematic diagrams illustrating that assaying closely linked target sequences in cf-DNA from maternal blood amplifies the digital representation of fetal sequences relative to maternal sequences. [0026] Figures 2A and 2B show a schematic diagrams illustrating that assaying closely linked target sequences in cf-DNA from maternal blood amplifies the digital representation of fetal sequences relative to maternal sequences.

[0027] Figures 3A and 3B show a schematic diagrams illustrating that assaying closely linked target sequences in cf-DNA from maternal blood amplifies the digital representation of fetal sequences relative to maternal sequences.

[0028] Figures 4A and 4B show a schematic diagrams illustrating that assaying closely linked target sequences in cf-DNA from matemal blood amplifies the digital representation of fetal sequences relative to maternal sequences.

[0029] Figures 5A and 5B show a schematic diagram illustrating that assaying closely linked target sequences in cf-DNA from maternal blood amplifies the digital representation of fetal sequences relative to maternal sequences.

[0030] Figure 6 shows an exemplary amplification curve showing positive reactions when maternal cf-DNA is assayed at 10 closely-linked target sequences.

[0031 ] Figure 7 illustrates that three linked target sequences (A, B1 and B2) can all lie on a "long" fragment (bottom) but only one can lie on a short fragment (top).

[0032] Figure 8 illustrates assay of four target sequences in a chromosomes.

[0033] Figure 9 shows an exemplary pattern of target sequences.

[0034] Figure 10 shows coincidence detection signals corresponding to

Chromosome 21 and SRY fragments contained within each reaction volume on the

DID chip.

DETAILED DESCRIPTION

Abbreviations and Definitions

[0035] D = diploid

[0036] T = triploid

[0037] D21 , D18, etc. = diploidy of chromosome 21 , diploidy of chromosome 18, etc.

[0038] T21 , T18, etc. = trisomy of chromosome 21 , trisomy of chromosome 18, etc.

[0039] cf-DNA = circulating cell free DNA.

[0040] fetal cf-DNA = circulating cell free fetal DNA (including DNA derived from the fetus, placenta, umbilicus, amniotic fluid, or other fetal-associated tissue. The transmission of fetal genomic sequences to the maternal circulation is incompletely understood, and as used herein, is not limited to a particular source. It is believed that most fetal cf-DNA is from the placenta.

[0041] fetal DNA = unless otherwise indicated "fetal DNA" is used interchangeably with fetal cf-DNA.

[0042] maternal cf-DNA = circulating cell free DNA of maternal origin.

[0043] maternal DNA = unless otherwise indicated "maternal DNA" is used interchangeably with maternal cf-DNA.

[0044] maternal blood (plasma, serum) = blood (plasma, serum) from a pregnant woman.

[0045] chr = chromosome (e.g., chr21 means chromosome 21 ).

[0046] Study chromosome = the chromosome for which aneuploidy, duplication or deletion is being studied.

[0047] Reference chromosome = a chromosome for which ploidy is known or assumed to which the study chromosome is compared. A reference chromosome is most often an autosomal chromosome assumed to be diploid.

[0048] Target sequence = A human chromosomal DNA sequence the presence or absence of which is detected. The length of the target sequence is the entire length required for detection by whatever detection system is being used. For example, if detection is by PCR amplification in combination with a molecular beacon probe the target sequence would be delimited by the position of the PCR primers.

[0049] Assay = The process of testing for the presence of a specific target sequence using digital analysis. For example, testing for the presence of any one or more of three different target sequences in a sample would comprise three assays.

[0050] Template molecule = A cf-DNA molecule that contains at least one target sequence.

I. Introduction for Assay of Closely Linked Targets in Fetal Diagnosis

[0051] A euploid human fetal cell contains 22 pairs of autosomal chromosomes. The ratio of one autosomal chromosome (e.g., chr21 ) to another (e.g., chr18) is 2:2. This 2:2 ratio is also reflected in the representation chromosome-specific sequences in fetal cf-DNA. That is, a chromosome 21 sequence will be found at about the same frequency as a sequence of another autosomal chromosome. In contrast, in a cell of a fetus trisomic for chromosome 21 , the ratio of chromosome 21 to another autosomal chromosome is 3:2, and chr21 sequences are over-represented in fetal cf-DNA by a ratio of about 3:2. In principle, fetal trisomy can be detected based on this overrepresentation.

[0052] However, because the maternal contribution to cf-DNA is usually 10 to 30- fold that of the fetus, and because most fetal DNA sequences cannot be distinguished from maternal sequences, this difference is difficult to reliably detect even using digital PCR. For example, Quake et al., supra, calculates that of fetal cf- DNA is 3% of total circulating cell free DNA, the ratio of representation of a chr21 sequences (both maternal and fetal) to sequence of another chromosome is 2.03:2.00.

[0053] The present invention provides new method for detecting fetal aneuploidy and other genetic abnormalities using digital PCR, which provides higher sensitivity than alternative methods, requires less starting material, and is more economical.

[0054] Importantly, although the following discussion focuses on chromosome 21 trisomy, this is merely to simplify and clarify the discussion. The method of the invention is broadly applicable to other fetal aneuploidies, other fetal genetic lesions, and other diagnostic or screening assays in which circulating cell free-DNA is detected, as is discussed below. In particular, applications in oncology are discussed below.

II. Overview of Assay of Closely Linked Targets in Fetal Diagnosis

[0055] The present invention is based on part on the observation that the size distribution of fetal cf-DNA differs from that of maternal cf DNA. As noted above, most fetal cf-DNA (hereinafter, "fetal DNA") is less than 300 b long, while most maternal cf-DNA (hereinafter "maternal DNA") is larger, often greater than 1 kb in length. Li et al., supra. This size difference has previously been the basis for certain methods for enrichment of fetal cf-DNA from plasma. See, e.g., Li et al., 2006, Recent developments in the detection of fetal single gene differences in maternal plasma and the role of size fractionation Ann N Y Acad Sci 1092:285-92.

[0056] In one aspect of the present invention, probes are used to detect DNA sequences of interest. Generally, the presence, absence or representation of the sequences of interest provides diagnostic or prognostic information to a patient or prospective parent. Each sequence of interest may be referred to as a "target sequence." Each interrogation of a DNA sample for the presence, absence or representation of an individual target sequence is referred to as an "assay." Thus, using multiplexing methods, multiple "assays" can be carried out simultaneously in a single reaction volume.

[0057] In the method, a DNA sample containing fetal and maternal DNA (or more generally a DNA sample containing a distinct DNA populations having different size distributions) is partitioned into many aliquots, or reaction volumes. As used herein, the term "partition" is not limited to any particular mechanism of distributing portions of a sample into numerous discrete reaction volumes. The distribution is usually adjusted, taking into account reaction volume size, DNA concentration, number of target sequences, and the number of aliquots used, so that most but not all reaction volumes contain at least one copy of at least one target sequence. In some embodiments, between about 10% and about 90% of the reaction volumes assayed provide a positive signal (indicating the presence of at least one target sequence in the aliquot). Often between about 20% and about 80% of the reaction volumes assayed provide a positive signal. In experiments in which more than 10^3,reaction volumes are assays, the proportion of reaction volumes with positive signals may be less than 10%. Preferably the total number of targets ("imputed targets" i.e. DNA fragments with one or more target sequences) will be >1000.

[0058] Methods for calculating the optimal representation per aliquot of a target sequence are provided in Dube et al., 2008, "Mathematical analysis of copy number variation in a DNA sample using digital PCR on a nanofluidic device," PLoSONE, 3, e2876. In general, an amount of DNA equal to that found in about 0.1 to about 10000 microliters maternal plasma is partitioned into about 500 - 10,000 reaction volumes. For example, DNA from 1 ml maternal plasma can be partitioned into about 3000 reaction volumes.

[0059] The optimal amount of plasma/DNA can be determined empirically for a given platform (e.g., number and volume of reaction volumes, number of assays, etc.). For an exemplary calculation we can assume (i) about 500 genome equivalents DNA/ ml plasma = 1000 copies of each chromosome (of which about 10% is fetal DNA); (ii) 10 target sequences are assayed per chromosome, each 50 base pairs long and with a gap of 50 base pairs between assays. Under these conditions the multiplier of fetal signals by ten-plexing equals about 5x (see Example 2, infra) resulting in about (100 x 5 =) 500 imputed targets. The multiplier for total cf- DNA is about 2.5, resulting in about (1000 x 2.5 =) about 2500 imputed targets (total). The multiplier for maternal cf-DNA is about 2.2, so maternal DNA results in about (900 x 2.2 =) 1980 imputed targets. (Imputed targets are the most probable number of fragments with target sequences). DNA from 100 ul of plasma in one panel would result in about 250 imputed targets per panel, which is on the lower end of the ideal DNA concentration.

[0060] In one embodiment on average each reaction volume contains about 0.2 to about 10 DNA fragments comprising one or more target sequences. Lower amounts of DNA may be used by increasing the number of assays.

[0061 ] The particular method used to determine whether a target sequence is or is not present in an individual reaction volume is not critical, as is discussed below, so long as a single copy of the target can be detected. For example, in one approach the target sequence is amplified using the polymerase chain reaction (PCR), and the amplicons, if present, are detected using a fluorescent probe. Detection of a target sequence in a reaction volume can be referred to as a "positive reaction" or "positive signal."

[0062] In one illustrative aspect, the present method involves partitioning cf-DNA and assaying each reaction volume for a plurality of closely linked target sequences. By assaying multiple (e.g., 10) closely linked sequences, the number of reaction volumes in which positive signals result from the presence of fetal cf-DNA is multiplied relative to positive signals resulting from maternal cf-DNA. Because both maternal and fetal DNA is fragmented, and as a consequence of the difference in size distribution between fetal and maternal molecules, this provides a significantly better readout and allows an accurate characterization of fetal genotype to be obtained using fewer reactions. Alternatively, for a given number of reactions, the method results in higher confidence in the results of the analysis. The differential increase in signal (i.e., between maternal and fetal sequences) is a consequence of the difference in size between (smaller) fetal cf-DNA and (larger) maternal cf-DNA. Closely linked target sequences are likely to be present on a single molecule of maternal cf-DNA, but unlikely to be linked on a single molecule of fetal cf-DNA. Targets on a single large DNA molecule will partitioned into a single reaction volume, resulting in a single positive signal, while the same target sequences dispersed on a number "N" of short DNA molecules will be distributed to "N" reaction volumes, resulting in "N" positive signals. Although the magnitude of signal from the single large DNA molecule may exceed that of any individual smaller DNA, the smaller DNAs will account for a greater number of reaction volumes with a positive signal.

[0063] This can be illustrated using a series of highly schematic figures. In these figures, each large rectangle represents an approximately 1.5 kilobase maternal cf- DNA molecule segregated in a single reaction volume, and each black rectangle represents a fetal cf-DNA molecule of 150 bases segregated in a single reaction volume. The numbers 1 -9 represent nine closely linked target sequences. Figure 1 A illustrates two maternal molecules and 18 fetal molecules. As illustrated by the alignment of the rectangles, each set of nine fetal molecules, in aggregate, contain the same target sequences as found on a maternal molecule. Figure 1 A illustrates the case in which both the fetal and maternal cells are diploid for the chromosome containing target sequences 1 -9. Figure 1 B illustrates the case in which the fetal cells are triploid for this chromosome, resulting in an increased number of fetal molecules relative to the diploid state.

[0064] (As noted above, the figures are schematic and those of skill will immediately recognize several respects in which the representations deviate from actual sample of cf-DNA. For example, cf-DNA will contain a heterogeneous mixture of fragments of a variety of lengths, rather than molecules of uniform length and with termini (break points) in register. In addition, as noted above, maternal cf-DNA is present in maternal plasma in considerable excess, not shown in the illustration.) [0065] Figures 2A and 2B illustrates that an assay that detects "Target 6" would result in a different number of positive signals euploid and triploid cases (i.e., 4 positives in the euploid case vs. 5 positives in the triploid case). Prior proposed methods attempt to measure this increase in positives. Figures 3A and 3B illustrate that by multiplexing more than one target on the chromosome may be assayed. In this case, Targets 6 and 16 are assayed for. The diploid case results in 8 positive signals and the triploid case represents 10 positive signals. Although there is no differential increase in signal from the fetal cf-DNA (10/8 = 5/4) assaying multiple targets in this fashion has the effect of reducing the total amount of cf-DNA required to detect a difference between the diploid and triploid cases.

[0066] Figures 4A and 4B illustrates that by assaying multiple closely linked sequences, the signal from fetal cf-DNA relative to signal from maternal cf-DNA is increased. Figure 4A shows a diploid case in which there are 20 positive signals (2 from maternal sequences and 18 from fetal sequences) and Figure 4B shows a triploid case in which there are 29 positive signals (2 from maternal sequences and 27 from fetal sequences).

[0067] Figures 5A and 5B illustrate assaying a reference chromosome to aid in assessing the increased number of positive reaction volumes in the triploid case. Figure 5A illustrates assaying multiple closely linked targets (A-J) in a reference chromosome sequence known or assumed to be diploid in the fetus. Figure 5A shows that in the diploid case, the number of positive reactions is similar for the (diploid) test chromosome and the (diploid) reference chromosome (20:20). Figure 5B illustrates that in the diploid case the number of positive reactions for the (triploid) test chromosome exceeds the number for the (diploid) reference chromosome (29:20).

[0068] Thus, in one aspect the invention relates to amplifying signal from fetal cf- DNA by assaying for multiple closely linked target sequences. In one aspect, the invention provides a method for diagnosing fetal chromosomal aneuploidy of a specified chromosome or chromosome region by (a) providing a sample comprising circulating cell-free DNA (cf-DNA) obtained from human maternal plasma; (b) partitioning the sample into a plurality of reaction volumes, where on average each reaction volume contains cf-DNA in an amount equal to that found in about 0.02 to about 10 microliters plasma; (c) assaying each reaction volume for the presence of each of two or more DNA target sequences, where the target sequences are closely linked in the specified chromosome or chromosome region; (d) determining the number of reaction volumes that contain at least one DNA molecule containing at least one of the two or more of DNA target sequences; and (e) comparing said number of reaction volumes with a reference number characteristic of a known chromosome ploidy, wherein a number of said reaction volume that is higher or lower than the reference number is indicative of fetal chromosomal aneuploidy of the specified chromosome or chromosome region.

[0069] Certain aspects of the invention are discussed In greater detail below. III. Introduction to Coincidence Detection

[0070] The present invention may be referred to as "Coincidence Detection." Coincidence Detection is useful for genetic analysis of a first population of genomic DNA fragments that is mixed with a second population of genomic DNA fragments, where the size distribution of fragments differs between the two populations. The invention provides methods for detecting the presence, absence or representation of DNA sequences of interest in a specified population of fragments (e.g., the population of shorter fragments) without, or with reduced, background noise from the fragments of the other population. One important example of such a mixed population is the cell-free circulating DNA (cf-DNA) in plasma of a pregnant woman which contains smaller fragments of fetal origin and larger fragments of maternal origin. A second important example of a mixed population is cell free DNA from patients with certain cancers. In some such cancers DNA fragments from apoptotic cancer cells are smaller that cell-free DNA from other cells. In others, the DNA fragments from cancer cells are larger. The pattern characteristic of specific cancers can be determined by reference to the scientific literature. Thus, the invention finds application in prenatal diagnosis and cancer diagnosis and prognosis, for example. However, the method of the invention is broadly applicable to any diagnostic or screening assays of a DNA sample having at least two distinct size populations, and particularly DNA of two or more origins which is circulating cell free-DNA in blood. Distinct size populations includes combinations of populations in which the predominant size ranges overlap, but for which the median size of fragments differs. For example a first population in which >95% of the fragments are in the range 50 bp-500 bp is distinct from a second population in which >95% of the fragments are in the range 400 bp-5 kbp.

[0071 ] Although the Coincidence Detection method is broadly applicable, for the sake of clarity and to simplify the discussion, the description below focuses on detection of fetal aneuploidies and other fetal genetic properties such as point mutations, which may or may not be associated with morbidity, polymorphisms (including single nucleotide polymorphism referred to as SNPs), small insertions or deletions, and the like. Genetic variations other than aneuploidy are referred to herein as "mutations".

[0072] It will be apparent from the discussion below that the distinction between detecting aneuploidy and detecting mutations such as SNPs is made solely to simplify the discussion and thereby for clarity. Both determinations may be carried out using the methods of the invention, simultaneously and on the same DNA sample (e.g. in the same reaction volumes).

[0073] In some embodiments the method can be used to detect fetal aneuploidy. The aneuploidy can be of an entire chromosome (e.g., T21 ) or aneuploidy of a chromosome region sometimes referred to as "partial aneuploidy" (e.g., a duplication or deletion). Aneuploidy of a specified chromosome is detected based on relative representation of the chromosomal sequences in fetal cf-DNA compared to representation of a different chromosomal sequence of known ploidy.

[0074] The normal ratio of one autosomal chromosome (e.g., chr21) to another (e.g., chrl ) is 2:2. This 2:2 ratio is also reflected in the representation chromosome- specific sequences in fetal cf-DNA. That is, a chromosome 21 sequence will be found at about the same frequency as a sequence of another autosomal chromosome. In contrast, in a cell of a fetus trisomic for chromosome 21 , the ratio of a chromosome 21 sequence to sequence of another autosomal chromosome is 3:2, and chr21 sequences are over-represented in fetal cf-DNA by a ratio of about 3:2. In principle, fetal trisomy can be detected based on this overrepresentation.

[0075] However, because the maternal contribution to cf-DNA is usually 10 to 30- fold that of the fetus, and because most fetal DNA sequences cannot be distinguished from maternal sequences, this difference is difficult to reliably detect even using powerful methods such as digital PCR. The Coincidence Detection method makes it is possible to determine representation of fetal sequences without high background from the maternal sequences in the sample and/or to amplify the relative signal of fetal sequences relative to maternal sequences in the sample. By subtracting most maternal signal and retaining most fetal signal an accurate determination of fetal ploidy is possible.

[0076] In some embodiments the Coincidence Detection method can be used to detect the presence or absence of a specific sequence (i.e., mutation) in fetal cf- DNA. Certain genetic mutations or polymorphisms, or combinations of mutations or polymorphisms, are correlated with certain phenotypes or correlated with likelihood of developing a disease, or are otherwise of interest. For example, a fetus that is homozygous for a mutation in the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) gene will develop cystic fibrosis. However it may be impossible to distinguish, with currently used methods, the fetal sequence from a maternal sequence. For example, if the mother is heterozygous for the mutation, it is generally not possible to determine the fetal genotype. Using the present method it is possible to determine fetal sequence(s) without interference from the maternal sequences in the sample. [0077] A large number of disease-associated variations are known, including, for example, variations associated with cystic fibrosis, Canavan disease, Familial dysautosomia, sickle cell anemia, Tay Sachs disease, thalassemia, Huntington's Disease, and others. However, the term "mutation," as used herein, does not imply a detrimental characteristic. For example, if two prospective parents are both heterozygous for a morbidity-associated allele, the methods of the invention may be used to determine that a fetus is homozygous for the normal allele.

IV. Overview of Coincidence Detection Method

[0078] The presence, absence or representation of DNA sequences of interest in the population of fragments being studied provides diagnostic or prognostic information to a patient or prospective parent. Examples of sequences of interest include sites of known mutations with diagnostic or prognostic value, sites of single nucleotide polymorphisms, sequences that are generally conserved (i.e., sequences that vary only rarely from individual to individual), and others. In the Coincidence Detection method sequences of interest in one population (e.g., fetal DNA) are distinguished from similar or identical sequences in the second population (e.g., maternal DNA) by (i) partitioning the mixed DNA sample into numerous small reaction volumes and (ii) detecting "target sequences" present in the reaction volumes. Target sequences are short (typically less than 100 bp in length) sequences in genomic DNA. In the Coincidence Detection method each target sequence is closely linked in the genome to at least one other target sequence, as is discussed in more detail below.

[0079] In the method, the DNA sample containing, for example, fetal and maternal DNA (or more generally a DNA sample containing a distinct DNA populations having different size distributions) is partitioned into many aliquots, or reaction volumes. As used herein, the term "partition" is not limited to any particular mechanism of distributing portions of a sample into numerous discrete reaction volumes (several exemplary approaches are described below). The distribution is usually adjusted, taking into account reaction volume size, DNA concentration, number of target sequences and the number of aliquots used, so that most, preferably almost all, reaction volumes contain one or fewer copies of at least one target sequence. For example, in one embodiment on average each reaction volume contains about 0.01 to about 0.5 DNA fragments comprising one or more target sequences, often about 0.01 to about 0.3, about 0.01 to about 0.10, about 0.05 to about 0.3 or about 0.05 to about 0.3. In some embodiments, the average is selected to minimize the number of reaction chambers with more than target-sequence-containing DNA fragment, and may be lower, particularly when very large numbers of reaction volumes art tested. The optimal amount of a DNA sample to be distributed to a predetermined number of reaction volumes can be determined empirically, for example, by using a dilution series. Alternatively the amount can be calculated based on the mass of DNA (in which the distribution may assume that 90% of the mass of DNA is maternal and 10% fetal) a measured. Alternatively the amount can be specified as a proportion of the DNA obtained from a specified volume of plasma (e.g., 2 mis) using a specified isolation method. Other methods and approaches will be evident to those of skill in the art.

[0080] The genomic DNA fragments present in each reaction volume are then assayed for the presence of a defined set of target sequences, for example by using quantitative PCR methods. At least two, and usually several, different target sequences are assayed for, using multiplex methods. The term "assay," as used herein, refers to interrogation for the presence, absence or representation of a specific target sequence. Thus, interrogating a DNA sample using three different probes that detect (e.g., hybridize to or amplify) three different target sequences is referred to as three "assays" even though the detection is carried out on a single DNA sample. Thus the process of genetic analysis of a sample from a patient comprises at least two, and usually several, "assays" and is referred to as a "clinical test."

[0081] Target sequences may be detected using a variety of art-known methods, some of which are described below. Typically target sequences are detected using molecular probes that hybridize to and/or amplify the target sequence and, directly or indirectly, produce a detectable signal such as a fluorescence at a characteristic wavelength. For example, in one approach the target sequence is amplified using the polymerase chain reaction (PCR), and the amplicons, if present, are detected using a fluorescent probe. In another approach melting probes are used, often in combination with asymmetric PCR, to detect a target. See Seipp et al., 2007, "Unlabeled oligonucleotides as internal temperature controls for genotyping by amplicon melting," J Mol Diagn. 9:284-9; Liew et al., 2007, "Closed-tube SNP genotyping without labeled probes/a comparison between unlabeled probe and amplicon melting," Am J Clin Pathol. 127:341 -8.

[0082] In a clinical test each reaction volume is assayed for the same set of target sequences. Generally the DNA sample is combined with the reagents necessary for sequence detection (e.g., probes, primers, enzymes, etc.) prior to partition or distribution of the sample into distinct reaction volumes. However, it is also possible to partition the DNA sample and then add all or some of the necessary reagents. This can be accomplished by fusing droplettes containing sample with droplettes containing reagents, by depositing reagents in reaction chambers (e.g., prespotting), wells of a multiwell plate, or the like prior to distribution of the DNA sample, by using microfluidic methods to add reagents to aliquots of sample, or the like. Detection of one or more target sequences in a reaction volume can be referred to as a "positive reaction volume" or a "positive signal."

[0083] The relative positions of target sequences and the signals (e.g., probe labels) used in their detection are important elements of the method. In the Coincidence Detection method, assays are carried out for "closely linked" target sequences, i.e., sequences close to each other in the genome. Two target sequences that are linked in the chromosome are more likely to be physically linked on a fragment of genomic DNA than are sequences separated by a great distance. Target sequences are selected so that at least certain combinations of target sequences are likely to be found on a "long" DNA fragment and unlikely to be found on a "short" DNA fragment. For example, a pair of linked target sequences be separated by about 300 bp cannot be found on a single 250 bp DNA fragment, but may be found on a single 600 bp DNA fragment. If most fetal cf-DNA fragments at shorter than 250 bp and most maternal cf-DNA fragments are larger than 250 bp, a fragment with only one of the target sequences is more likely to be a fetal cf-DNA than is a fragment with both of the target sequences. Conversely, a fragment with both of the target sequences is more likely to be a maternal cf-DNA than is a target sequence with only one of the target sequences. In the Coincidence Detection method at least some closely linked target sequences are detected in a manner that allows the linked targets to be distinguished from each other. For example, two linked target sequences may be detected by hybridizing one target sequence with a probe labeled with the dye VIC and the other with a probe labeled with the dye FAM. [0084] The positions and nature of target sequences can vary, as described elsewhere herein. For purposes of illustrating the Coincidence Detection approach to determining the representation of fetal target sequences with little or no "noise" from maternal target sequences in fetal diagnostics, we will assume that each target sequence is 50 base pairs long that there is a gap of 200 base pairs between neighboring target sequences. (Target sequence are "neighboring" target sequences when no other target sequence being assayed for lies between them on the chromosome.)

[0085] By assaying for linked target sequences and selecting different labels for the respective detection reagents (e.g., probes and/or primers) it is possible to distinguish between targets on short DNA fragments those on longer fragments. For example, when a DNA sample is partitioned so that most reaction volumes contain zero or one fragment from the chromosome region in which the linked target sequences are located, a long fragment is more likely to have two linked target sequences and a short fragment is more likely to have only one of the linked target sequences. For example, Figure 7 illustrates that three linked target sequences (A, B1 and B2) can all lie on a "long" fragment (bottom) but only one can lie on a short fragment (top).

[0086] Assuming a hypothetical case in which a reaction volume contains one and only one polynucleotide from this region of the chromosome, if each of A, B1 and B2 is detected in a single reaction volume the fragment must be at least 550 bp in length while if only target sequence A is detected the fragment must be less than 550 bp in length. If each of A plus B1 but not B2, or A plus B2 but not B1 is detected in a single reaction volume the fragment must be at least 300 bp in length. If B1 and B2 are labeled so they cannot be distinguished, detection of A plus B1 and/or B2 indicates a fragment is at least 300 bp in length. It will be appreciated that by selecting (i) the number, position, length and spacing of target sequences and (ii) the combination of labels (or other distinguishing feature) used to detect each target sequence it is possible to deduce information about the size of the fragment in each reaction volume. The practitioner can elect to record results (number of fragments, presence or absence of specific DNA sequences) for only "short" (however defined) fragments, for only "long" (however defined) fragments, or for both.

[0087] For illustration, if "short" fragments are less than 300 bp and "long" fragments are at least 300 bp, target sequence A is detected using a VIC-labeled probe (green) and target sequences B1 and B2 are detected using FAM-labeled probes (blue), reaction volumes containing only a short DNA fragment can be identified as those that emit green but not blue. Conversely, reaction volumes emitting blue and green can be considered "not short." See Table 1 .

Table 1

[0088] In Rows 1 -3 of Table 1 , the double or triple hit could be a consequence of having two or three short fragments in a single reaction volume rather than one longer fragment. By counting this as "not short" as in this example, the number of "short" fragments containing target sequence A could be slightly underestimated. Note also that Rows 1 -3 cannot be distinguished because B1 and B2 are detected with the same label.

[0089] In some embodiments, where detection of multiple signals occurs, the multiple signals can be said, for conceptual purposes, to cancel each other out or null by coincidence (i.e., coincidence detection) and can be ignored for purposes of analyzing sequences of short (e.g., fetal cf-DNA) polynucleotides. By ignoring, for purposes of analysis, any signal from a reaction volume positive for both A and at least one of B1 and B2 it is possible to limit the sequence analysis to only short, typically fetal, cf-DNA fragments.

[0090] Rows 5 and 6 in Table 1 are denoted "not informative" because the fragments could be as short as 50 bp, or could be on very long (e.g., 5 kb) fragments in which target sequence B1 or target sequence B2 is near the end of the fragment. It should be noted that in practice, in some embodiments the "B" only signals can be counted as "short" based on the statistical determination that this population while containing some long fragments will be enriched for short fragments. (Also see Table 4, infra.)

[0091] In the example illustrated above (Table 1 and Figure 7), the probe that detects target sequence A is distinguishable (i.e., differently labeled) from the probes that detect targets B1 and B2. Although there are several advantages to using such differential labeling, the method can also be carried out without differential labeling by detecting differences in intensity of signal, for example. Referring to Table 1 , signal from a reaction volume containing a long DNA fragment will be two or three times the intensity of signal from a short DNA fragment. In a different approach melting probes are used and target sequences are distinguished by differences in melting temperature.

[0092] Referring again to the example illustrated above, the probes that detects target sequences B1 and B2 are not distinguishable (i.e., differently labeled). By using three labels additional information can be obtained, as illustrated in Table 2. In this illustration, target B2 of Figure 7 is detected with a differently labeled probe, denoted "C". In the hypothetical example of Table 2, "short fragments" are those less than 400 bp and "long" fragments are this more than 500 bp.

Table 2

[0093] In the two rows denoted "ambiguous" the fragment is at least 300 bp and is more likely to be short than fragments corresponding to Row 1 , and less likely to be short than fragments corresponding to Row 2. The fragments could be as short as 300 bp, or could be on very long (e.g., 5 kb) fragments having a terminus (or breakpoint) that lies between B1 and A, or between A and C on the chromosome. As discussed above, clinical tests may be designed to fit a particular need by varying several parameters (e.g., target sequence position and length) one of which is the definition of "long" and "short" fragments. The "ambiguous" reaction volumes in Table 2 can be counted as likely long, likely short, or ignored (not counted) depending on the application.

[0094] The following nomenclature is used herein to describe embodiments in which target sequences are distinguishable (detected with a differently labeled probe or otherwise distinguishable): Target sequences identified by different letters (e.g., A, B) are detected with differently labeled and distinguishable probes. For example target sequence A (or signal A) is distinguishable from target sequence B (or signal B). Target sequences identified by the same letter and different numbers (e.g., B1 and B2) may or may not be differently labeled. Often they are not differently labeled, e.g., to reduce the number of different labels required for a single clinical test. However, conceptually target sequences identified by the same letter and different numbers could be differently labeled, and if practical, such labeling may provide additional information. Table 3 below provides exemplary labels. Many others are known in the art. In addition to labeling probes, an additional dye may be included as a passive reference.

Table 3

[0095] In Figure 7, target sequence A is an "internal" target sequence and target sequences B1 and B2 are "flanking" target sequences. An internal target sequence is a target sequence situated between two other closely linked target sequences being assayed. A flanking target sequence is one that does not lie between two other closely linked target sequences. (Note that a flanking target sequence may lie between two target sequences being assayed if only one is closely linked. For example a flanking target sequence may lie between two target sequences being assayed, one of which is separated from the flanking target sequence by 200 bases (i.e., is closely linked) and the second of which is separated from the flanking target sequence by 5kb (i.e., is not closely linked). As is discussed above, Table 1 illustrates that detection of an "internal target sequence" only reaction volume (e.g., A-only) is more informative than detection of a "flanking target sequence only" (e.g., B1 only, or B2 only) reaction volume

[0096] As is discussed below, in certain embodiments multiple "internal" target sequences are assayed. The multiple internal target sequences may be a series of internal closely linked target sequences (I) with a pair of flanking target sequences (F), e.g., "F1 -I1 -I2-I3-I4-I5-F2." Alternatively they may be a number of individual internal target sequences with flanking sequences, e.g., "F1 -I1 -F2" + "F3-I2-F4", etc.

[0097] The approaches discussed above used detection of at least three closely linked target sequences. By counting reaction volumes positive for "internal-only" target sequences it is possible to identify signal from short fragments with a high degree of confidence. However, it is also possible to conduct the assay using only two closely linked sequences, e.g., A-B.

[0098] With such an approach reaction volumes positive for both target sequences are deemed noninformative or long. This group includes most long fragments (e.g., most maternal cf-DNAs) and a small number of short fragments. Reaction volumes in which only one target sequence is found are counted as short. These reaction volumes include some with long fragments (in which the target sequence lies near the end of the fragment) but are enriched for short fragments. See Table 4, in which both A and B can be considered "flanking sequences." Table 4

[0099] In some embodiments several pairs of sequences with each member of the pair being closely linked to the other.

[0100] Although counting internal target sequences and not counting flanking target sequences allows you to more completely exclude "long" fragments, counting reaction volumes positive for flanking sequence-only is also informative and makes it possible to increase the number of positive reaction volumes. The positive reaction volumes can evaluated using known statistical methods.

[0101] Table 5 illustrates an embodiment in which both internal and flanking sequences are assayed and both are counted.

Table 5

[0102] In some embodiments positive reaction volumes positive for an internal target sequence but no flanking target sequences (e.g., Row 4 of Table 5) and reaction volumes positive for a flanking target sequence but not internal target sequence(s) (e.g., Rows 5 and 6 of Table 5) are detected are enumerated separately and subject to separate statistical analysis (for example, giving greater weight to internal target sequences). If desired the internal-target-sequence-only positive reactions of a study chromosome or region can be compared to the internal-target- sequence-only positive reactions of a reference chromosome or region and, likewise, the flanking-target-sequence-only positive reactions of a study chromosome or region can be compared to the flanking-target-sequence-only positive reactions of a reference chromosome or region.

[0103] The invention may practiced using a variety of patterns of target sequence spaced to allow short and long fragments to be distinguished (by observation or statistical analysis). For example and not limitation, target sequences can be assayed as ABC, AB1 B2, AB1 A, C1 ABC2, ABCD, (AB)_n, C(AB)_nC, AB_nA where N 2. By counting the reaction volumes positive for a single label, and deeming reaction volumes with multiple labels noninformative, it is possible limit the count to essentially only reaction volumes for which the positive signal is from short fragments.

[0104] In one embodiment the pattern of target sequences is (AB)_n (n > 3; usually n = 3-25, often 3-10, and sometimes 5-10). This approach can be carried out using only two probe labels. Although internal and flanking target sequences are not distinguished, the ratio of internal sequence to flanking sequences can be high, providing a high level of confidence in distinguishing long and short sequences. See Table 6. By adding flanking sequences, C(AB)_nC, and scoring A-only and B-only as short, a somewhat higher level of confidence is possible, but at the cost of an additional label.

Table 6

[0105] In some embodiments of the present invention multiple groups of closely linked target sequences probes, such as two, three, four, and more. In some embodiments multiple sets of three closely linked sequences in a chromosome are assayed. For example, the pattern of target sequences can be represented as below, illustrating assays of four groups of three target sequences, one of which is internal:

B1 -A1 -B2

B3-A2-B4

B5-A3-B6

B7-A4-B8

[0106] To identify reaction chambers in which the target sequence containing fragment is short, reaction chambers in which both A and B are detected can be ignored. In some embodiments, only internal sequences are counted, in which case only reaction chambers with A-signal only are deemed "short."

Terms "long", "short" and "closely linked"

[0107] It will be recognized that the parameters "long," "short," and "closely linked" are relative terms and the design of assays will be dictated by the particular application, DNA preparation methods, etc. For example, in the context of fetal diagnostics using conventional methods of isolating cf-DNA, long (maternal) fragments may be defined as at least 300 basepairs in length and short (fetal) fragments may be defined as less than 300 or less than 200 basepairs in length. However, in other contexts (e.g., detecting allelic imbalance in cancer; alternative methods for isolation of cf-DNA) the parameters "long" and "short" can vary. Given a first population of "long" fragments and a second population of "short" fragments in general, it is desirable that the range defining "short" encompasses at least 50%, preferably at least 70%, often at least 80% and sometimes at least 90% of the fragments in the second population, but excludes at least 50%, preferably at least 70%, often at least 80% and sometimes at least 90 of the fragments in the first population. Likewise, in general, it is desirable that the range defining "long" encompasses at least 50%, preferably at least 70%, often at least 80% and sometimes at least 90% or 95% or even about 100% of the fragments in the first population, but excludes at least 50%, preferably at least 70%, often at least 80% and sometimes at least 90% or 95% or even about 100% of the fragments in the second population.

[0108] Li et a!., 2004, "Size Separation of Circulatory DNA in Maternal Plasma Permits Ready Detection of Fetal DNA Polymorphisms," Clinical Chemistry 50: 1002- 1 1 , provide guidance about the sizes of fetal cf-DNA, although it will be recognized that purification methods, fetal age and other factors can affect the distribution. According to Li et al., about 70% of the fetal cf-DNA is less than 300 bp in length and more than about 90% is less than about 500 bp in length. In contrast, about 75% of the maternal cf-DNA is larger than 300 bp and about 40% is larger than 500 bp in length.

. In general the fetal fraction is considered to be DNA less than about 300 bp in length, sometimes less than 200 bp in length. In general the maternal fraction is considered to be DNA greater than 500 bases length, and more usually greater than 1 kb in length, such as 1 -23 kb in length. It will be appreciated that the "fetal fraction" less than 300 bases in length may contain maternal molecules, and the maternal fraction may contain some fetal molecules.

[0109] Similarly, two target sequences are considered "closely linked" when the length of the fragment required to detect both sequences (usually at least the length of the two target sequences plus the distance between the target sequences) is greater than the upper bound of the length defined as "short," equal to or greater than less than the lower bound of the length defined as "long," and small enough so that both closely linked sequences can be detected in a fragment in a size range that encompasses most of the long fragments in the DNA sample being analyzed. Thus, target sequences can be closely linked when the distance between them is small enough so that they are likely to be detected on a single long chromosomal DNA fragment present in the sample, and significantly less likely to be detectable on a single "short" chromosomal DNA fragment in the sample.

[0110] For example, consider a system in which short fragments are defined as those 50-299 bp in length and long fragments are defined as those at least 300 bp in length. Given two target sequences, each 50 bp in length, and separated from each other by 200 bp, either one, but not both, can be found on a short fragment and both could be found on a long fragment. This is also true if short fragments are those 50- 200 bp in length and long fragments are those at least 400 bp in length. [01 11] In addition to being spaced far enough apart so that closely linked sequences are usually not found on the same short fragment, closely linked sequences must be close enough to each other so that they can both be detected on a significant number of "long" fragments in the sample. For example, if "long" fragments in a population are empirically determined to be in the range of 400 bp to 20 kbp, but 90% of the long fragments are 800-1000 bp range, the closely linked target sequences should be selected so they would be detectable on an 800 bp fragment. That is, for purposes of this invention, target sequences 2 kb apart would not be closely linked. Similarly, if "short" fragments in a population are empirically determined to be in the range of 25 bp to 600 bp, but 90% of the short fragments are 100-200 bp range, the closely linked target sequences should be selected so only one would be detectable on an 200 bp fragment. That is, in this hypothetical and assuming target sequences are 50 bp in length, two closely linked target sequences would be at least 101 bp apart (50 + 50 + 101 < 200) and not more than 700 bp apart (50 + 50 +700 < 800).

Ploidy Reference Values

[0112] In some embodiments of the invention (e.g., embodiments in which aneuploidy is to be determined) the number of positive reaction volumes can be compared to a reference value characteristic of a known or assumed chromosome ploidy. For example clinical test for aneuploidy of chromosome 21 the number of reaction volumes positive when assayed for Chr21 target sequences can be compared to a reference value characteristic of diploidy (determined under defined conditions of DNA concentration, etc.). A measured value that is the same as the reference value is indicative of diploidy and a measured value that is significantly higher or lower than the reference value is indicative of aneuploidy (e.g., triploidy). The reference value may be determined by assaying a reference chromosome or reference sequence in parallel with the assays of the study chromosome. In "parallel" with means that each reaction volume is assayed (usually simultaneously) for both the target sequences in the study chromosome or region and for those in the reference chromosome or region. Alternatively, the reference value may be determined by distributing a portion of the DNA sample into a different set of reaction volumes, and assaying independently. For example, using the BioMark™ 12.765 Digital Array, which partitions a sample into 12 panels of 765 reaction volumes, a study chromosome may be assayed in one panel and a reference chromosome may be assayed in a different panel of the same device. In one embodiment the ratios of "long" to "short" of each panel are compared. Reference values are discussed in greater detail herein below.

[0113] Tables 7 and 8 illustrate determination of aneuploidy of chromosome 21 using chromosome 1 as a reference. The values in the tables are hypothetical. In this example, chromosome 21 is assayed at three closely linked targets (B1_, A and B2, where A is an internal target sequence and BJ. and B2 are flanking target sequences). Chromosome 1 is assayed at three closely linked targets (B3, C and B4, where C is an internal target sequence and B3 and B4 are flanking target sequences). The pattern of target sequences can be represented as follows:

Ch21 B1 -A-B2

Ch1 B3-C-B4

In this example, A is detected with a probe labeled with VIC, C is detected with a probe labeled with FAM, and B1 , B2, B3 and B4 are detected with probes labeled with Cyber Green.

Table 7

- Long 21 180

+ + NOT INFORMATIVE

- Short 21 31

- + SHORT

+ Long 1 175

+ - NOT INFORMATIVE

+ Short 1 20

- - SHORT

[0114] The Table 7 results are consistent with a D21 fetus because the number of A only chambers is about equal to that of C only chambers. The Table 8 results are consistent with a T21 fetus because the ratio of A only chambers to C only chambers is about 3:2.

[0115] In an other example, the chromosome 1 and chromosome 21 flanking target sequences are detected using distinguishable probes. The pattern of target sequences can be represented as follows:

Ch21 B1 -A-B2

Ch1 D1 -C-D2

A, C, B1 and B2 are detected using probes labeled as above. D1 and D2 are labeled with CAL Fluor orange and Texas Red, for example.

Table 9

Table 10

[0116] Note that, as presented, Tables 9-12 illustrate clinical tests in which fragments containing chromosome 1 target sequences and which fragments containing chromosome 21 target sequences are not found in the same reaction chamber. In fact, this segregation of target sequences is not required.

[0117] In yet another example, each chromosome is assayed at 2 rather than 3 target sequences. The pattern of target sequences can be represented as follows:

Ch21 B-A

Ch1 D-C

A, C, B1 and D2 are detected using probes labeled as above.

Table 1 1

Table 12

[0118] As discussed above, in some embodiments more than 3 linked target sequences can be assayed and a variety of labeling schemes used. Figure 8 illustrates assay of four target sequences in a chromosomes, where the pattern of target sequences can be represented as Ch21 A-B1 -B2-B3. In this example, a probe for A is labeled with FAM and probes for B1 , B2 and B3 are labeled with VIC.

[01 19] In this approach, reaction volumes positive for target sequences detected as both FAM (A) and VIC (B1 ), and optionally B2 or B2 and B3) are not counted as indicative of long fragments, and reaction volumes positive for only FAM or VIC are counted as short fragments. It will be recognized that some reaction chambers positive for one signal only may represent long fragments. However, as illustrated in Figure 8, for a given genome equivalent, detection of B1 , B2 and B3 from a small fragment population will result in three positive reaction volumes per haploid genome equivalent, while detection of B1 , B2 and B3 from a large fragment population will result in one positive reaction volumes per haploid genome equivalent. In the case of prenatal diagnostics using cf-DNA the effect of this multiplier is decreased somewhat because the maternal (long) fragments are present in excess over the fetal (short) fragments; however, the amplification of signal attributable to fetal cf- DNA is dramatic and provides effective diagnostic methods. By subtracting a proportion of the long fragments from the analysis the power of this approach is increased even further.

Mutation analysis

[0120] In another embodiment, the techniques of the present invention are used to detect a mutation and differentiate between mutations present or absent on short versus long fragments. For example, in some clinical tests a sample is probed for a specific mutation, such as a point mutation associated with disease or single nucleotide polymorphism (SNP). A large variety of methods for SNP and mutation analysis are well known and can be used in the Coincidence Detection method. Exemplary approaches are discussed below. Typically a mutation is detected using two probes (or sets of probes) one of which detects one sequence (e.g., one allele) and the other of which detects an alternative sequence (e.g., the alternative allele) and labeled so that the two target sequences can be distinguished.

[0121 ] In one exemplary embodiment, illustrated in Figure 9 the pattern of target sequences can be represented as C1 -X-C2 where X can be A or B, and A and B are different alleles (or A is normal and B is mutant, etc.).

[0122] Upon analysis, detection of probe A binding will indicate the wild-type short fragment. In contrast, detection of probe B binding will indicate a mutated fragment. Similarly, if binding of probes B and C represent a coincident event, then the fragment contained within the reaction volume can be recorded as a long fragment having the same mutation. Or, if binding of probes A and C are detected in a coincident event, then the fragment will be recorded as a long wildtype fragment. This differential analysis can be useful in particular for analyzing SNPs or mutations on short versus longer fragments, such as in analysis of hereditary from the mother to the fetus in the context of maternal plasma samples, or differential detection of mutations on fragments from cancerous versus healthy/non-cancerous cells. For illustration, Tables 13 and 14 show hypothetical results for clinical tests for two combinations of fetal and maternal genotypes. Note that the results can be analyzed without reference to the long (maternal) sequences. However, because the maternal genotype can be separately determined (e.g., from a DNA source not containing fetal sequences) the "maternal" results provide a powerful method for validation.

Table 13: Fetus and Mother Both Heterozygous

Table 14: Fetus Homozygous Mother Heterozygous

This approach can be particularly in mutation analysis, discussed below. In one approach each internal sequence is an allele of a different mutation. Any initial screen can be done in which all "normal" alleles are detected using a single label, with secondary screening used if the presence of any "non-normal" (but unidentified) alleles are detected.

[0123] In some embodiments the number of target-sequence containing fragments per reaction chamber can be increased to significantly greater than 0.3. For example, in an assay for 3 mutations each of which is distinguished by a differently labeled probe up to three target-sequence-containing fragments (if different fragments) could be isolated in a single reaction chamber. Similarly, in an assay with for multiple unlinked groups of target sequences, having >0.3 fragments per reaction volume may increase the number of chambers deemed "not informative" but A-only chambers, for example, would still be accurate.

[0124] Certain aspects of the invention are discussed In greater detail below.

V. Collection and Preparation of Circulating Cell-Free DNA

[0125] Methods for isolation of circulating cell-free DNA from maternal blood are well described in the literature. See, e.g., Li et al., 2004, Clinical Chemistry 50: 1002- 1 1 . Typically about 2-20 ml peripheral blood is collected and, optionally, processed to remove cells. Preferably the sample is processed to obtain plasma, although it is also possible to use serum or blood. Cell-free DNA may be isolated from plasma or serum using well known methods. Because of the heightened sensitivity of the assays of the invention, less maternal blood is needed to obtain a statistically reliable result. For example, the methods can be carried out using 0.05-10 mis plasma, more usually 0.1 -5 mis plasma.

[0126] Care should be taken to preserve the physiological size distribution of DNA fragments (i.e., not to further fragment DNA) during processing.

[0127] Although blood (e.g., plasma or serum) is the preferred source of cf-DNA, any source containing maternal and fetal cell-free nucleic acids may be used. For example, DNA-containing body fluids such as urine, pancreatic fluid, spinal fluid and lymph may be used.

[0128] Optionally the cf-DNA may be enriched for molecules derived from the fetal cells. Generally, enrichment is based on the size difference between fetal and maternal cf-DNA molecules. Methods for enrichment include separating DNA by size include gel electrophoresis (see Li et al., supra), size exclusion chromatography, differential precipitation, microcapillary or microchannel separation (see, e.g., Foquet et al., 2002, DNA Fragment Sizing by Single Molecule Detection in Submicrometer Sized Closed Microfluidic Channels, Analytical Chemistry 74:1415-22), and other methods. In general the fetal fraction is considered to be DNA less than about 500 bases in length, sometimes less than 300 bases in length. In general the maternal fraction is considered to be DNA greater than 500 bases length, and more usually greater than 1 kb in length, such as 1 -23 kb in length. It will be appreciated that the "fetal fraction" less than 300 bases in length may contain maternal molecules, and the maternal fraction may contain some fetal molecules. For purposes of size selection the fetal enriched fraction can be considered to be DNA less than about 500 bp in length. See Li et al., 2004, "Size Separation of Circulatory DNA in Maternal Plasma Permits Ready Detection of Fetal DNA Polymorphisms," Clinical Chemistry 50:1002-1 1 .

[0129] In one approach a sample is enriched for short (e.g. <300 bp) DNA fragments by preferentially precipitating small molecules and separating out longer DNA strands. For blood samples, raw plasma, plasma lysate, or DNA extracted from plasma can be mixed with -1 -5% PEG 8000 (MW 7000-9000) at ~1 M NaCI (or other monovalent salts). The sample may be incubated at 4 °C on ice (at -20 °C) for 1 hr or longer (overnight). To isolate the short DNA fragments from the long genomic DNA strands, the sample can be centrifuged for a period of time (e.g., at -1500 g or higher, 1 -60 minutes) and the supernatant is removed. The supernatant will contain the short DNA fragments, and the larger molecular weight species will be localized in the pellet. Polyethylene glycol can also be used to enrich long DNA fragments, e.g., by selecting the long fragment-rich pellet rather than the supernatant. Sample enrichment using polyethylene glycol based precipitation. For example, raw plasma, plasma lysate, or DNA extracted from plasma can be mixed with -1 -5% PEG 8000 (MW 7000-9000) at -1 M NaCI (or other monovalent salt). The mixture is incubated at about -20 - 4°C for 1 hr or longer (e.g., overnight). To enrich for short DNA fragments the sample is centrifuged at -1500 g (or higher) for 1 -60 minutes, after which the supernatant, containing short DNA fragments, is separated from the pellet, containing larger MW species precipitated by the PEG.

[0130] In some embodiments, prior to distribution of the DNA sample into various reaction volumes, the DNA is denatured (e.g., thermally) to single stranded form. When detection methods are used that can detect a target sequence present in single stranded DNA (e.g., PCR-based methods) this will effectively double the number of copies of detectable target sequences.

VI. Digital Analysis

[0131 ] "Digital analysis" refers to a method in which a nucleic acid sample is distributed into many separate reaction volumes so that most reaction volumes have at most a small number of template molecules, detecting the individual template molecules, and correlating the number of reaction volumes containing template molecules with the representation of template in the starting sample. For a general review see Zimmerman et al., 2008, "Digital PCR: a powerful new tool for noninvasive prenatal diagnosis?" Prenatal Diagnosis published online DOI:10.1002/pd.2150. When the individual template molecules are detected using PCR, the digital analysis method can be referred to as "Digital PCR" (see Vogelstein and Kinzler, 1999, PNAS 96:9236-41 ; Pohl and Shih, 2004, "Principle and applications of digital PCR" Expert Rev Mol Diagn 4(1 )). Analogously, when individual template molecules are detected using the ligase chain reaction (LCR), the method can be referred to as "Digital LCR."

[0132] In one approach a DNA sample is diluted into multiwell plates so that there is, on average, 0.5 template molecules per well, combined with PCR amplification reagents, and amplifying sequences from individual template molecules using PCR. See Vogelstein and Kinzler, 1999, PNAS 96:9236-41 . The proportion of wells with positive signals can be correlated with the representation of the template in the starting material.

[0133] A related method involves distributing a DNA solution (containing reagents sufficient for PCR amplification of target sequences) into one or more microfluidic channels, and partitioning the channel(s) into numerous isolated reaction volumes ("massive partitioning"). A microfluidic device adapted for this method is described in U.S. Pat. App. Pub. US 2005/0019792, incorporated herein by reference. Such a device is sold by Fluidigm Corp. (S. San Francisco CA) as the BioMark™ 12.765 Digital Array which partitions a sample into 12 panels of 765 reaction volumes using integrated fluidic circuit (IFC) valves to partition the sample. See, e.g., Qin et al., "Studying copy number variations using a nanofluidic platform" Nucleic Acids Research 2008:1 -8 and Zimmerman et al., 2008, supra. However, other microfluidic devices and valve-types can be used in the present invention. In a related embodiment a device is used in which a DNA solution is introduced into a network of channels and partitioned into a plurality of reaction volumes. In some embodiments the number of reaction volumes is more than 1000, more than 5000, more than 9000 or more than 20,000. In some embodiments the liquid volume of each reaction volume is less than 100 nanoliter (nl), and is usually in the range of 0.1 to 10 nl.

[0134] Another approach to digital PCR is described in Dressman et al., 2003, PNAS 100:8817, incorporated herein by reference; U.S. Pat. App. Pub 07/0065823 (emulsion PCR). Digital analysis can be carried out using a variety of platforms including BioMark™ 12.765 Digital Array (Fluidigm); OpenArray (BioTrove), multi- well (e.g., 96-, 384-, or1536-well) plates; SmartChip™ Real-Time PCR System (Wafergen) and others.

VII. DNA Distribution

As noted above, the distribution of the DNA sample is usually adjusted, taking into account reaction volume size, DNA concentration, number of target sequences, and the number of aliquots used, so that most reaction volumes contain zero or one copy of a target sequence-containing DNA fragment. In some embodiments, between about 1 % and about 30%, more preferably 20%, of the reaction volumes assayed provide a positive signal (indicating the presence of at least one target sequence in the aliquot). In experiments in which more than 10^3,reaction volumes are assays, the proportion of reaction volumes with positive signals may be less than 1 %. Preferably the total number of targets ("imputed targets" i.e. DNA fragments with one or more target sequences) will be >1000. In some embodiments, an amount of DNA equal to that found in about 0.1 to about 10000 microliters maternal plasma is partitioned into about 500 - 10,000 reaction volumes. For example, DNA from 1 ml maternal plasma can be partitioned into about 3000 reaction volumes.

VIII. Target Detection

[0135] As alluded to above, a number of approaches can be used to detect individual target sequence present in a single reaction volume. In general there is an amplification of the target sequence, or of a probe(s) that recognize the target sequence, and the "amplicons" are detected. Methods such as PCR, ligase-chain reaction, nucleic acid sequence-based amplification (NASBA), transcription- mediated amplification (TMA), Invader assay, rolling circle amplification, multiplex ligation-dependent probe amplification and other methods can be used in the present invention by selecting primers and/or probes specific for the closely-linked target sequences of the test chromosome(s) and reference chromosome(s) are provided to each reaction volume. Edwards et al. 1994, "Multiplex PCR: advantages, development, and applications," PCR Methods Appl. 3:S65-75; Elnifro et al., 2000, "Multiplex PCR: Optimization and Application in Diagnostic Virology" Clin. Microbiol. Rev. 13: 559-570; Fang et al., 1995, "Simultaneous analysis of mutant and normal alleles for multiple cystic fibrosis mutations by the ligase chain reaction" Hum Mutat. 1995;6:144-51 P; Schouten et al., 2002, "Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification". Nucleic Acids Res. 30: e57; Wu and Wallace, 1989, Genomics 4:560; Landegren et al., 1988, Science 241 :1077; Kwoh et al., 1989, Proc. Natl. Acad. Sci. USA 86:1 173; Guatelli et al., 1990, Proc. Nat. Acad. Sci. USA, 87:1874; Sooknanan et al., 1995, BioTechnology 13: 563-65. Each of the foregoing references are incorporated herein by reference in their entirety for all purposes. Reagents for carrying out the amplification/detection will vary with the method(s) selected, but can be easily determined by those of skill in the art. Typically the reagents are combined with the DNA-containing sample prior to partitioning the sample into separate reaction volumes.

[0136] One approach involves PCR amplification of the target sequences followed by detection of the resulting amplicons using, for example, Molecular Beacons (see e.g., Piatek, 1998, Nat. Biotechnol. 16:359-63 and Tyagi et al., 1998, Nat. Biotechnol. 16:49-53.

[0137] In another approach the ligase detection reaction is used. This involves ligation of oligonucleotides that bind to the target directly adjacent to each other and PCR amplification of the ligation product. In this manner it is possible to introduce tags into the ligation product (flanking the target sequence) and to use these as primer binding sites (e.g. all ligation products for chromosome 21 have the same tags).

[0138] In another embodiment the ligase chain reaction (LCR) is used. LCR is a nucleic acid amplification method that uses 2 pairs of complimentary probes. Probes 1 and 2 are designed to anneal to target DNA immediately adjacent to one another. The 'nick' between them is recognized by DNA ligase and ligated, so that 2 oligomers of e.g. 25 nt become one 50 nt oligo. The mixture is then heated so that the probe and target DNA are separated. On cooling, further copies of Probes 1 and 2 can anneal to the target and Probes 3 and 4 can anneal to the ligation product of probes 1 and 2, formed in the prior round.

[0139] In another approach an Invader assay (Third Wave Technologies, Madison, Wis.) is used. In the Invader system there is a linear amplification of signal when a target sequence is present, but there is no amplification of the target sequence itself. An oligonucleotide, designated the signal probe, is complementary to the target sequence. A second oligonucleotide, designated the Invader Oligo, contains the same sequence. The Invader Oligo interferes with the binding of the signal probe to the target nucleic acid such that the 5' end of the signal probe forms a "flap." This complex is recognized by a structure specific endonuclease, called the Cleavase enzyme. Cleavase cleaves the 5' flap of the nucleotides. The released flap binds with a third probe bearing FRET labels, thereby forming another duplex structure recognized by the Cleavase enzyme. This time the enzyme cleaves a fluorophore away from a quencher and produces a fluorescent signal, signal probe can be designed to hybridize with either the reference (wild type) allele or the variant (mutant) allele. Further details sufficient to guide one of ordinary skill in the art is provided by, for example, Neri, et al., 2000, Advances in Nucleic Acid and Protein Analysis 3826:1 17-25).

[0140] Amplified sequences can be detected using any number of sequences including detection based on fluorescence resonance energy transfer (FRET), bioluminescence resonance energy transfer (BRET), time-resolved energy transfer (TRET) and the like (see Tyagi et al., 1996, "Molecular Beacons: Probes That Fluoresce Upon Hybridization" Nature Biotechnology 14: 303-308; Heid et al., Genome Research 6:986 994 (1996); Gibson, U. E. M, et al., Genome Research 6:995 1001 (1996); Holland et al., Proc. Natl. Acad. Sci. USA 88:7276 7280, (1991 ); U.S. Pat. No. 5,210,015 to Gelfand, U.S. Pat. No. 5,538,848 to Livak, et al., and U.S. Pat. No. 5,863,736 to Haaland). Amplification and/or detection of multiple closely linked target sequences can be carried out using well known "multiplexing" methods.

[0141 ] In typical multiplexing applications differently labeled probes are selected so that each unique target sequence can be separately detected. In the present method, it is necessary to distinguish the target sequences from the study chromosome from the target sequences of the reference chromosome, but distinguishing or individually detecting the closely linked target sequences on the same chromosome is not required. Usually, to reduce cost and the number of reagents required one label (e.g., fluorophore per chromosome is used). For example, VIC might be used as a label for multiple different assays of study chromosome target sequences and FAM might be used as a label for multiple assays of a reference chromosome. Alternatively, in some embodiments the methods of the invention (in which digital signal from fetal sequences is amplified) can be combined with methods for distinguishing maternal and fetal alleles. See, e.g., Pohl and Shih, 2004, supra.

[0142] Similarly, although quantitative amplification methods (e.g., real time quantitative PCR) can be used, quantitation is not required in digital analysis because each reaction volume is assayed for the presence or absence of one copy of the target. Quantitative methods may be used in some cases; for example quantitative methods may be used to confirm whether or not a single reaction volume contains multiple copies of a target sequence.

[0143] In a related approach, PCR amplification is used to distinguish short and long fragments. In one embodiment a pair of forward primers (F1 and F2) and one reverse primer (R) are used. F2 and R are closely linked and can be used to generate an amplicon from a short fragment, while F1 and R are separated by a greater distance and can be used to generate an amplicon from a long fragment but not a short fragment. The production of longer and shorter amplicons can be detected using probes, or by sequencing of the generated amplicons. Using the sequencing approach, it can be determined whether specified target sequences are on the same or different amplicons.

[0144] Other detection approaches may be used. For example, analysis by probe melting may be used. See Seipp et al., 2007, "Unlabeled oligonucleotides as internal temperature controls for genotyping by amplicon melting," J Mol Diagn. 9:284-9; Liew et al., 2007, "Closed-tube SNP genotyping without labeled probes/a comparison between unlabeled probe and amplicon melting," Am J Clin Pathol. 127:341 -8.

IX. Target sequences

[0145] As detailed above, the methods of the invention involve assaying each reaction volume for the presence of each of target sequences that are closely linked on specified chromosome. [0146] A target sequence is a short DNA sequence (e.g., less than 200 bases, more often less than 100 bases, even more often less than 50 bases, and sometimes less than 25 bases) found in the genome being analyzed (e.g., the human genome). In one embodiment, a target sequence is unique in the genome. That is, when the amplification and/or detection method selected is applied to the human genome, a single site is amplified/detected over a threshold level. In one embodiment, detection means (e.g., probes and primers) are selected so that a single probe and/or primer set, for example, can be used to detect multiple different closely linked target sequences. For example, a probe using inosine at selected positions could be used to detect non-identical sequences. In general, any detection means may be used to detect any plurality of target sequences, so long as only the closely linked target sequences are detected.

[0147] It will be appreciated that PCR amplification and detection of amplicons may involve primers and probes that hybridize to three different sites (two primer biding sequences and a probe binding sequence between them). However, the chromosomal sequence corresponding to each unique amplicon is considered a single target. That is, PCR amplification of two adjacent sequences, using a total of four PCR primers, and subsequent detection of the amplicons using a total of two labeled probes would constitute assays of two target sequences.

[0148] In some embodiments, two or more sets of closely linked target sequences are identified on the same chromosome. For example, five closely-linked target sequences at one locus and five closely-linked target sequences at a different locus can be assayed. In some cases it is advantageous to assay at a single locus.

[0149] At least two, and typically at least three closely linked target sequences are assayed. However, more often at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, or at least 10 closely linked target sequences are detected. In some embodiments more than 10 closely linked target sequences are detected. In some embodiments the number of closely linked target sequences assayed is in the range of 8-20. In some embodiments the number of closely linked target sequences assayed is greater than 20, such as from 21 -100 target sequences.

[0150] In most applications, e.g., prenatal diagnoses, when two target sequences are "closely linked" the sequences are more likely to be encoded on a single polynucleotide molecule 1 kb in length than they are to be encoded on a single polynucleotide molecule 100-300 bases in length. In some embodiments the closely linked target sequences are within 1 kb of each other. In some embodiments the closely linked target sequences are within 0.2 or 0.3 kb of each other. Closely linked target sequences can be identified by reference to the published human genome and the scientific literature. Two target sequences are "closely linked" when the sequences are more likely to be encoded on a single polynucleotide molecule 1 -10 kb in length than they are to be encoded on a single polynucleotide molecule 100- 300 bases in length. Closely linked target sequences are usually within 2 kb of each other. In some embodiments the closely linked target sequences are within 1 kb of each other. In some embodiments the closely linked target sequences are within 0.5 kb of each other. Closely linked target sequences can be identified by reference to the published human genome and the scientific literature.

[0151 ] For illustration. Tables 15 and 16 provide PCR primer pairs that amplify closely linked target sequences on Chromosome 21 and Chromosome 18, respectively. Each primer pair amplified an approximately 50 nucleotide target sequence, and there is an approximately 50 nucleotide gap between each target sequence. Primer pairs below can be selected to identify target sequences with various spacing. For example, using 21 .01 in combination with 21 .03 defined two 50 bp target sequences separated by 50 bp separated by 50, spanning 150.

Table 15

Forward primer Reverse primer

#21.01 TTTCTCTCG G TTTG G AG AG G T ATT CTC TTC CAG AAA CGC CTG

#21.02 GACAGCAAAACCTCATTTTTTAACC GTG CCT GCT TTT GAC GTT G

#21.03 G CAAAG AAG CG G CCAAC GAA CCC TGC TTT CAA GAA GAA GT

#21.04 TGTTTTGTTAAAGGGGGTGG CCC GCA AAA TCC ACA CTG A

#21.05 AGGAGAGATCATTCGTATTCGACC TGA GGT CAG AAG TCC TCA TTT ACG

#21.06 CATGCCACTTTCTCCTGGATT ACT AAC TGG ACA GTG GAC GGT TT

#21.07 GACCTGAAGTAGCATTTAGTTACCAAG CCT GTG TGG AGT GGG CTG T

#21.08 CTG CAG CCTG CCTTCTCTC CAC CCT TGT CAG CGC AAT

#21.09 CATTTCTGTACCAGCTCCCCTA GGG AAA AGA CAC ACA AGA CAA CA #21.10 CCTTTGGATGTCCTCAACAAATAA AAT ACT GCA TGA CAA TAA ATA TTT GTC C Table 16

Forward primer Reverse primer

#18.01 GGAGAACCCATGTCAAAAACAA CTTTGACCTTTGGTGGTGTCTTT

#18.02 CGGTCCTTATGGTGTCCGT G C AG CT A AG GTC AG AG CTG AAA #18.03 CCACTCCCTGGAACCTTCTT ACTC A A AG GAAGAAGCCAAGTGT #18.04 CCCAACCTCCTTGGTGTTTC AGATGGTGATGTTCGTCTATGGA #18.05 GTGAAATGTGCTGTAAGCTTTATGAG AC ACG AG A A AC A A ATCCTCTG C

#18.06 CCTGCCACATAGTAGGAGTCAACA CAGGTGGTATCTACATTTAGTGATCAAT #18.07 AG AG G G CTG AGTAG GTACCTCAG G CCTGTCTTCACAG CACTCTG

#18.08 GATGTTCTTTCTGACCAGTGCTTT GACCTCAGCCACCTTCCAA

#18.09 GGTATGGCTTGTATGGTCTTGG G G AT A AG AG G G CTTG G GTCTG

#18.10 G CTCTGTATTCTTGG G GTATTTTG TGTTAGTCAAGATAGGCCAATACAGA

[0152] In some embodiments, at least two, at least three, at least 4, at least 5, or at least 6, at least 7, at least 8, at least 9, or at least 10 or more target sequences are assayed, where each of the target sequences is closely linked to each of the others. For example, 5 such target sequences can lie within a single 2 kb region of a study chromosome.

[0153] In alternative embodiments several groups of closely linked targets are assayed, where each target is closely linked to other targets in the groups, but are not necessarily closely linked to targets in other groups. For example, at least two, at least three, at least 4, or at least 5 target sequence may be closely linked to each other in one region of the study chromosome, and at least two, at least three, at least 4, or at least 5 target sequence may be closely linked to each other in a different region of the chromosome. In various embodiments there may be 10 assays distributed as five to 10 groups with at least two closely linked targets in each group, four groups with four closely linked targets in each group, and the like.

[0154] Any number of algorithms (including computer implemented versions) are known in the art for designing primers and/or probes to amplify or detect a particular sequence.

[0155] It will be understood that the methods of the invention can be used to detect any aneuploidy including trisomies, such as trisomy of chromosome 8 (Warkany syndrome 2), 9, 13 (Patau syndrome), 15, 16, 18 (Edwards syndrome), 21 (Down syndrome), 22 (Cat eye syndrome), sex chromosome trisomy (e.g., XXY; XYY), sex chromosome tetrasomy and pentasomy; monosomies such as monosomy of chromosome 4 (Wolf-Hirschhorn syndrome), 7 (Williams syndrome), 1 1 (Jacobsen syndrome), 17 (Miller-Dieker syndrome/Smith-Magenis syndrome), 22 (22q1 1 .2 deletion syndrome), X (Turner syndrome), as well as partial deletions, mosaicism and the like. In some embodiments the genetic lesion is partial monosomy or partial trisomy (loss or gain of a part of a chromosome) such as partial monosomy of chromosome 5 (Cri du chat syndrome). In the case of partial aneuploidy target sequences on a reference chromosome can be used as described above. For example, Cri du chat syndrome is due to a partial deletion of the short arm of chromosome 5 (e.g., 5p15.2-3 monosomy) and can be detected according to the methods of the invention by assaying closely linked target sequences in that region (study sequences) compared to closely linked sequences in other areas of chromosome 5 (reference sequences).

[0156] More than one study chromosome can be assayed for aneuploidy or other genetic lesion, each with a set of closely linked target sequences. In general, detection reagents are selected so that sequences from the different study chromosomes can be distinguished. For example, molecular beacons labeled with different fluorophores can be used to distinguish sequences from different chromosomes (or if desired, different loci on a single chromosome, or, if desired, individual closely linked targets).

X. Reference Numbers

[0157] As discussed above, aneuploidy of an entire chromosome or aneuploidy of a chromosome region (e.g., a duplication or deletion, sometimes referred to as "partial aneuploidy") can be determined by comparing positive reactions obtained by assaying multiple closely linked target sequences in the chromosome or chromosome region and comparing that number with the number of positive reactions expected if the chromosome or chromosomal region was euploid (e.g., diploid). For example, when assaying for target sequences on chromosome 21 , the number of positive reactions from fetal cf-DNA will be greater in the case of a T21 fetus compared to a D21 fetus.

[0158] In one embodiment a reference number characteristic of a known chromosome ploidy is determined by statistical analysis of a large number of samples, under defined reaction conditions and DNA concentrations. The results from any assay of a study chromosome in a particular maternal blood sample can be compared to the statistically determined reference number or standard curve. [0159] However, the comparison is most often and most conveniently done by conducting digital analysis of target sequences on a reference chromosome (or chromosome region) of known (or assumed) ploidy from the same maternal blood sample. The assays of the reference chromosome are typically carried out simultaneously with the assays of the study chromosome and generally using the same partitioning device. For example, in test for chromosome 21 triploidy using a DNA sample, the same sample can be assayed (preferably simultaneously using multiplexing) for the number of positive reactions in an assay of chromosome 1 (if chromosome 1 is known or assumed to be diploid (in, for example, both the fetal and maternal cf-DNAs). Assays of study targets can be distinguished from studies of reference targets by using assays that provide a different signal for the reference and study chromosomes. When the number of positive reactions in assays for the study chromosome is higher or lower than the number of positive reactions in assays for the test chromosome it is an indication of fetal chromosomal aneuploidy of the specified chromosome or chromosome region. In a related embodiment, as discussed below partial aneuploidy can be detected by assaying target sequences in a first chromosome region (characterized by likelihood of a duplication or deletion) and comparing the number of positive reactions to those obtained when target sequences in a region of the chromosome not characterized by likelihood of a duplication or deletion.

[0160] The reference chromosome may be, and preferably is, also assayed at a plurality of closely linked target sequences, in the manner discussed above. In some embodiments the number and/or distribution of target sequences assayed for the reference chromosome is the same, or about the same, as the number and/or distribution of target sequences on the study chromosome. Thus, in some embodiments at least two, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, or at least 10 target sequences are assayed on both a study chromosome and a reference chromosome. In some embodiments several groups of closely linked targets are assayed on both the study and target chromosomes.

[0161 ] In some embodiments target sequences are assayed on two or more study chromosomes, each of which serves as a reference chromosome for the other. For example, in a simultaneous assay of chromosome 21 and chromosome 18 sequences a deviation from euploidy can be determined by comparing hits from the chr21 assays and the chr18 assays if either of the chromosomes is aneuploid. In this example, a deviation (from D21/D18) may not be detectable, from these assays alone, when the fetus is T21/T18, for example, or has monosomy of both chr21 and chr18. However, many such combinations will be so rare (e.g., a combination that prevents early development of the fetus) and/or will be readily identifiable by clinical (e.g., ultrasound) analysis, so as to not affect the interpretation of the assay. It will be appreciated that in a case in which three of more chromosomes are assayed, hits from each chromosome can be compared to each of the others to detect anomalies. It will also be appreciated that the principles discussed above equally apply, with obvious modifications, to assessment of partial aneuploidy in which assays are carried out on euploid and potentially aneuploid regions of the same chromosome.

XI. Other Applications

[0162] Although discussed thus far primarily in the context of prenatal analysis of the fetal genome, the methods of the application have broader application.

Oncology

[0163] Genetic instability, including aneuploidy, and deletions and duplications of specific chromosomal sequences ("allelic imbalance"), is a hallmark of many tumors. Tumors release significant amounts of genomic DNA into systemic circulation. Analysis of circulating cell-free DNA in patients diagnosed with, or suspected of having, cancer can be used to detect allelic imbalance in tumor derived sequences for screening, diagnostic and prognostic purposes. See Pohl and Shih, 2004, supra and Giasuddin et al., 2008, "Applications of free circulating nucleic acids in clinical medicine: recent advances, "Bangladesh Med Res Counc Bull. 34:26-32. However, because circulating cell-free DNA includes a mixture of both DNAs from neoplastic and non-neoplastic cells, allelic imbalance in tumor-derived DNA can be difficult to detect. See Su et al., 2004, "Human urine contains small, 150 to 250 nucleotide- sized, soluble DNA derived from the circulation and may be useful in the detection of colorectal cancer," J Mol Diagn. 6(2):101 -7; Liu et al., 1999, "Loss of heterozygosity in tumor cells requires re-evaluation: the data are biased by the size-dependent differential sensitivity of allele detection" FEBS Lett. 26;462(1 -2):121 -8; Brant et al., 2003, "Increased Plasma DNA Integrity in Cancer Patients", Cancer Research 63:3966-68. Notably, a size differential between tumor and non-tumor-derived DNAs is characteristic of at least some tumors. Depending on the specific cancer, neoplastic cf-DNA may be shorter or longer than non-neoplastic cf-DNA. Without intending to be bound by a particular mechanism, this is likely a consequence of the origin of neoplastic cf-DNA from apoptosis of cancer cells.

[0164] Methods of the present invention may be used to differentially amplify signal from shorter cf-DNA fragments. In the case in which neoplastic DNA fragments are on average shorter than non-neoplastic DNA the digital signal from tumor DNA is amplified, while when neoplastic DNA fragments are on average longer than nonneoplastic DNA the digital signal is diminished. In either case, the change in signal can be detected by assaying multiple closely linked target sequences.

[0165] Specific allelic variations are associated with individual tumor types and can be readily determined by reference to the scientific literature. See. e.g., Suhai et a., 2001 , "Frequent allelic imbalance at the ATM locus in DNA multiploid colorectal carcinomas" Oncogenomics 20:6095-6101 . By reference to the literature closely linked target sequences known to be amplified or deleted in a cancer ("study sequences") may be identified, as well as closely linked target sequences in a chromosome or chromosome region having normal copy number ("reference sequences"). In some embodiments the methods of the invention (in which digital signal from tumor sequences is increased or diminished) can be combined with methods for detecting other genomic changes (e.g., alleles or point mutations) that distinguish cancer and non-cancer cells.

[0166] Aberrant or differential DNA methylation has been identified in lung cancer patients (Russo et al., 2005, "Differential DNA hypermethylation of critical genes mediates the stage-specific tobacco smoke-induced neoplastic progression of lung cancer," Cancer Res 65:1664-9), maternal blood (Poon et al., 2002, "Differential DNA methylation between fetus and mother as a strategy for detecting fetal DNA in maternal plasma," Clin Chem. 48:35-41 ) as well as other conditions. See, e.g., Giasuddin et al., 2008, supra. Methylation-specific PCR can be used as the amplification step in the invention for detection of differences in methylation according to the methods of the invention. See, e.g., Cottrell et al., 2003, "Sensitive detection of DNA methylation," Ann N YAcad Sci. 983:120-30).

Other conditions

[0167] The methods of the invention are applicable to any application in which two or more populations of DNA molecules can be distinguished by the size of the DNA and, in particular, methods in which two population of circulating cell-free DNAs are analyzed.

[0168] Cf-DNAs are reported to be increased on a number of conditions, including, for example, pre-eclampsia, intrauterine growth retardation (IUGR), and post traumatic inflammation. See, e.g., Margraf et al., 2008., "Neutrophil-derived circulating free DNA (cf-DNA/NETs): a potential prognostic marker for posttraumatic development of inflammatory second hit and sepsis" Sfroc/ 30(4):352-8.

XII. Examples

Assay of Closely Linked Targets in Fetal Diagnosis Example 1 : Increased detection of fetal sequences by assaying closely linked target sequences

[0169] Three 2-ml samples of maternal plasma (plasma of a pregnant woman) were collected. DNA was purified from each 2 ml sample and concentrated in a final volume of 25 microliters elution buffer. 3.3 ul of DNA were added to PCR reagents including primers and probes for 1 , 8 or 14 assays (i.e., sufficient to amplify and detect 1 target sequence, or 8 or 14 closely linked target sequences, in the Y Chromosome) to a total volume of 9 ul. See Table 17.

Table 17

Forward Primer Reverse Primer

SRY.01 TCTTA A AGTCTG A AG AG AA ATG AC A AG A TCCTCCC ATTTTATTTC AG CAA

SRY.02 G CAAAAC AG G AGTAACAACTACAG GT TTTTTATTAATAGGTTCTGTGTACTATGCAAC

SRY.03 ATCCCGTAAAATATGAAAAGAGTTCA ACAAACCAACCAC ATTTG GTTC

SRY.04 TCC ATG A A A AG A A A A AT ATG A A A ACTG T CTAAACTTTAGTAGATAACTAGGATACAAATTACAGTTT

SRY.05 ATACAAAGTG G AAATATAGTTG G CTCA TTGCTATCCAGATACATGTTTTAACTTG

SRY.06 CAGGAAATCCCTTACAAGATGAGA TGTCAG GTTTTG AAATG GTATGTTT

SRY.07 TGTTACACTGTGTGAAAAAGTCAGATAC G G ATC ATTC AGTATCTG G CCTCTT

SRY.08 AGA TTA ATG GTT GCT AAG GAC TGG AT CAGTAACTCCCCACAACCTCTTT

SRY.09 GAGAGGTGATGGCTGCACA TCTG ACTCTTTG G TTC ACCATGT

SRY.10 AGAGAAACAGATTCCAGAAGCATCT CCGGTCACGACACAATGTC

SRY. ll AGGTTACTTACAACCTATTATCTTAATTGGTAAT TCATGAACTGTTAAGATAACCCTTCA

SRY.12 G G G TA A A A AG TC AG TTTTC ACTATG A A C AG ACAAA A ATTG CTGCCACTT

SRY.13 TGTTGTTGCTATGGTTAGGTCTTG CAGTAGTACCGGGGAACGAAG

SRY.14 GATGGAGAGCGGCGACTAG TG G ACG CTGTACCTCTCC ATA

[0170] The forward primers are tagged with a short 5' sequence (probe binding sequence) that allows binding of one LNA probe for all amplicons for the generation of PCR signal by 5'-exonuclease cleavage (Zhang et al., 2003, "A novel real-time quantitative PCR method using attached universal template probe" Nucleic Acids Res 2003;31 :e123), This design allows the duplex detection of shorter amplicons than conventional TaqMan design with a probe between the primers

[0171 ] The resulting nine reaction mix containing samples were pipetted into the sample loading wells of a digital PCR chip (BioMark™ 12.765 Digital Array, Fluidigm Corp.) and distributed automatically into 765 reactions per panel. PCR was performed under standard thermocycling conditions for 50 cycles. After PCR, reaction chambers with an increased amount of fluorescence were scored as positive and counted.

[0172] Counts per panel are listed in the following table for three maternal samples carrying a male fetus and one control plasma sample from an adult male. As shown in Table 18, the increase of detected fetal DNA fragments using 8 or 14 assays in comparison to a single assay is 4.8-fold and 6.2-fold respectively.

Table 18

sample n = 1 n = 8 n = 14

1 3 15 18

2 4 15 20

3 2 13 18

male control

plasma 106 218 245

[0173] The increase of counts using multiple neighboring assays is greater in the maternal DNA samples than for the male control sample. This greater signal is attributable to the presence of short fetal cf-DNA in maternal plasma. Example 2: Increased of counts of chromosome 21 signals using 10 neighboring amplicons

[0174] DNA was purified from six samples of 2 ml of maternal plasma and concentrated into 25 ul elution buffer for each sample. 3.2 ul of DNA were added to PCR reagents including primers and probes sufficient for detection of 10 target sequences on Chromosome 21 (10 assays) to a total volume of 48 ul. PCR primers as shown in Tables 17 and 18, supra, were used. The forward primers of each chromosome are tagged with a common short 5' sequence (probe binding sequence) that allows binding of one LNA probe per chromosome. The distinct probe sequences for chromosome 18 and chromosome 21 are labeled with different fluorescent dyes to allow specific detection of products per chromosome.

[0175] For each of the 48 ul reaction mixes, 8 ul was pipetted into each of six sample loading wells of a digital PCR chip. From each loading well 4.65 of the reaction mix was distributed automatically into a panel of 765 reaction volumes. The total DNA per sample tested was from 0.614 ml plasma in a total of 6 x 765 = 4590 reactions (0.134 ul/reaction volume).

[0176] Counts of positive reactions per panel were converted into "detected fragments" per panel using poisson based distribution statistics in the digital PCR software. Figure 6 shows an example of amplification curves of one panel. Red curves are from reactions scored as positives.

[0177] In Table 19, "Targets" refers to hits (i.e., positive reaction volumes) per panel. "Fragments" is the Table 5 shows data (summarized in Table 20) from 6 samples with 6 panels per sample. "Targets" shows positive reaction volumes per panel. "Fragments" is the most probable number of fragments in the panel (calculated according to Dube et al., 2008, supra).

Table 19

Table 20 counts total fragments total

CV CV

per panel counts per panel fragments average 331 5.8% 1987 440 7.5% 2642

median 329 4.8% 1971 455 6.7% 2731

Example 3: Analysis

[0178] From 0.6 ml of plasma we detected 1032 to 4577 fragments, equal to 1680 to 7460 fragments per ml plasma. Previously analyzed pregnancy plasma samples contained 300 to 1500 genome equivalents. Thus a singleplex assay would detect 600 to 3000 molecules. The increase of counts from total DNA using 10 closely- linked target sequences is about 2.5 fold.

[0179] The increase of counts using assays of 10 closely-linked target sequences is greater for maternal cf-DNA than for total (control) DNA. Table 21 is a synthesis of data from several experiments including those in Examples 1 and 2 above, when the number of positives detected in a multiplex assay (e.g., a 10-plex assay) are compared to the number in a single-plex assay, the number of positive reaction volumes ("counts") resulting from fetal targets detected increases by 5-fold, while the number of positive reaction volumes resulting from maternal targets detected increases by only 2.2-fold. In Table 21 "100%" is defined as total number of counts from maternal plasma using one assay.

Table 21

approximate

increase of

counts single counts from 10 counts by 10- assay linked assays plex

fetal 10% 50% 5 X total 100% 250% 2.5 X maternal (total - fetal) 90% 200% 2.2 X Coincidence Detection

Example 4. Detecting Chromosome 21 and Y Chromosome (SRY) Using

Coincidence Detection

I. Introduction

[0180] This example demonstrates that Coincidence Detection can be used to differentiate between short and long DNA fragments, i.e., fetal and maternal DNA, as set forth herein. Specifically, this example describes the use of coincidence detection on chip to analyze XYYY genomic DNA as well as chromosome 21 and SRY fragments of fetal and maternal DNA present in plasma samples from pregnant and normal subjects.

II. Experimental Procedure

B. Partitioning Sample and Detecting Target Sequences

[0181 ] A DID open-channel, microfluidic device similar to the BioMark™ 12.765 Digital Array was designed and fabricated for conducting the experiments described in the following example. The device had 12 panels, each having a flow input for a sample or assay mixture. Instrumentation controlled the input of fluid into the chip. The chip architecture in this example partitions 465 μΙ_ of fluid into a plurality of discrete reaction volumes.

[0182] DNA samples combined with amplification reagents and probes for quantitative PCR and were distributed into reaction volumes on the microfluidic device. DNA was XYYY genomic DNA, three plasma samples taken from pregnant subjects, and two control plasma samples taken from normal male subjects. DNA from plasma samples was obtained using standard DNA extraction methods. The XYYY DNA was purchased. The samples were sufficiently dilute so that on average each reaction volume contained only a small number (e.g., zero or one) of DNA fragments containing a target sequence.

[0183] Assays were carried out for two target sequences in Chr21 and, separately, for SRY using TaqMan® probes. The Chr21 target sequences were separated by 376 bp and were assayed with differently labeled TaqMan® probes "A" and "B". The SRY target sequences were separated by 300 bp and were assayed with differently labeled TaqMan® probes "A" and "B". It was predicted that both target sequences could be detected on "long" fragments but only one or the other could be predicted on "short" fragments. The qPCR reactions were continuously monitored by a dynamic array reader. Signal was output to a computer equipped to record the data sent from the reader.

III. Results

[0184] Figure 10 shows coincidence detection signals corresponding to

Chromosome 21 and SRY fragments contained within each reaction volume on the DID chip. As described above, the DNA was sufficiently diluted so that only one fragment copy would be present in each reaction volume (depicted by each box within the panels). Coincident events were indicated by red and blue squares within a single reaction volume. In contrast, non-coincident events were signified by either a red or blue signal, indicating that only one of the target sequences was detected (e.g., A or B).

[0185] The coincidence detection data provided a wealth of information including, for example, the number of fragments as well as the ratios of short (noncoincident reaction volumes) to long (coincident reaction volumes) fragments. Table 22 lists the relative percentage of non-coincident events present in each of the 2 panels. The target sequences were separated by 300 and 376 nucleotides for SRY and

Chromosome #21 , respectively.

Table 22

Percentage of Non-Coincident Events in gDNA, plasma DNA, and cff DNA.

% non-coincident signal.

[0186] In addition, the cff DNA concentrations measured on DID chips agreed with reported literature values, but the counts were low (sampling error). The coincidence detection experiments concluded that cff DNA was -20 - 50 genome equivalents/ml (GE/mL), total DNA was -350 - 1400 GE/mL, and cff DNA/total DNA was -3 % to 8 %. Notably, 30-50% of maternal DNA signals were eliminated by coincident events measured as blue and red together.

IV. Conclusions

[0187] The example described herein illustrates the use of the coincidence detection technique on-chip to distinguish between signal stemming from short versus longer fragments. In particular, the method can be used to quantify fetal cf- DNA on-chip; the quantification values agreed with literature results. The example also provides independent confirmation by Digital PCR that cf-DNA was fragmented as expected based on reports in the literature. Furthermore, large volume plasma extraction can be integrated well with this method, which for example should be able to extract 2 ml plasma into 5 μΙ DNA with minor modifications.

Example 5. Coincident and Non-Coincident Detection with Multiple Probe Distances

I. Introduction

[0188] The example demonstrates that the distance between probes in a

Coincidence Detection experiment can affect the accuracy in differentiating between short and long fragments.

II. Experimental Procedure

[0189] Sample processing, distribution of DNA fragments on-chip, and PCR reactions were performed in a similar manner as described above in Example 4. Genomic DNA was used.

[0190] Figure 8 illustrates the coincidence detection strategy used in this example. Coincidence detection was conducted with four different TaqMan® probes to detect amplicons of both long DNA fragments and short cff DNA fragments. The TaqMan® probes were designated as A (red), B-i (blue), B₂ (blue), and B₃ (blue). A and Bi were 308 bp apart, A and B₂ were 528 bp apart, and A and B₃ were 676 bp apart. Coincident events (A plus B) were indicative of a long fragment, while noncoincident events were indicative of a short (<300 bp) fragment.

III. Results

[0191 ] Table 23 illustrates the results obtained from using probes A, B-i , B₂₍ or B₃ individually or in combination as shown. Table 23

Measuring Co-localization of Signals on Genom

[0192] Notably, coincident events showing co-localization of two signals, for example, can be dependent on the distance between the probes. As shown in Table 23, the percentage of non-coincident events increases as distance between A and B is increased. For panels 5, 6, and 7, the increase of single color events, e.g., A or B only, corresponded roughly with the distance AB, which was 306, 528, and 676 bp for AB-i , AB₂, and AB₃, respectively. Distances were measured from the 5'-end of A to the 3'-end of each B. As shown in Table 23, when the distance between A and B₃ was 676 base pairs, A and B only events increased to 6.6 and 9.6 %, respectively. This result likely occurred because an increasing number of DNA fragments shorter than 676 bp were detected began to show up as having only one color in a respective reaction volume.

[0193] Table 23 illustrates one advantage of coincidence detection: the amount of coincident event background from long DNA fragments can be reduced to further enhance detection of short cff DNA fragments. For example, a distance of 306 base pairs with ΑΒ_Ί provided fewer noncoincident signals than the 676 base pair distance of AB₃.

[0194] The probability of detecting a coincident event in a reaction volume depends on the distance between the probes and the length of the fragment. The following equation can be used to determine the probability:

P = 1 - (d)/(l) where p is probability of measuring coincidence of two signals (e.g., A & B) in one reaction volume, d is the distance between the outer ends of probes A and B, and I is the DNA fragment length. For example, if the fragment length is one nucleosome (150 bp), then when d is 306, the probability of coincidence signal is essentially zero. The following examples further illustrate use of the above equation in estimating probability of coincidence:

where I = 150 and d = 150 - p = 1/150

where I = 150 and d = 100 -» p = 33%

where I = 500 and d = 300 -» p = 40%

where I = 500 and d = 150 p = 70%

[0195] It is understood that the examples and embodiments described herein are for illustrative purposes only and that various modifications or changes in light thereof will be suggested to persons skilled in the art and are to be included within the spirit and purview of this application and scope of the appended claims. All publications, patents, and patent applications cited herein are incorporated by reference in their entirety for all purposes to the same extent as if each individual publication, patent or patent application were specifically and individually indicated to be so incorporated by reference. Citation of a publication is not to be construed as an indication that the publication is prior art.

Claims

WHAT IS CLAIMED IS:

1 . A method for detecting chromosomal aneuploidy of a specified chromosome or chromosome region comprising

a) providing a sample comprising cell-free DNA (cf-DNA) obtained from a biological fluid of a subject,

b) partitioning the sample into a plurality of reaction volumes and assaying each reaction volume for the presence of each of two or more DNA target sequences, wherein each of said two or more DNA target sequences is closely linked to at least one other of said two or more DNA target sequences in the specified chromosome or chromosome region, wherein the sample is partitioned so that most reaction volumes comprise not more than 10 discrete DNA fragments comprising a target sequence,

c) determining the number of reaction volumes that contain at least one DNA molecule comprising at least one of said two or more of DNA target sequences, d) comparing said number of reaction volumes with a reference number characteristic of a known or assumed chromosome ploidy,

wherein when said number of reaction volumes is higher or lower than the reference number it is indicative of chromosomal aneuploidy of the specified chromosome or chromosome region.

2. The method of claim 1 wherein the biological fluid is plasma.

3. The method of claim 2 wherein the subject is a pregnant woman and the presence or absence of a chromosomal aneuploidy of a specified chromosome or chromosome region of the fetus is determined.

4. The method of claim 1 wherein the subject is an individual diagnosed with or suspected of having cancer.

5. The method of claim 1 wherein the sample is partitioned so that most reaction volumes comprise not more than 5 discrete DNA fragments comprising a target sequence.

6. The method of claim 1 wherein between about 10% and about 90% of the reaction volumes assayed provide a positive signal indicating the presence of at least one target sequence in the reaction volume.

7. The method of claim 6 wherein on average each reaction volume contains about 0.2 to about 10 DNA fragments comprising one or more target sequences.

8. The method of claim 3 wherein on average each reaction volume contains cf- DNA in an amount equal to that found in about 0.02 to about 10 microliters plasma.

9. The method of claim 1 wherein the reference number characteristic of a known chromosome ploidy is determined by a simultaneous assay of multiple target sequences on a reference chromosome that is other then the specified chromosome, or on a chromosome region that is other then the specified chromosome region.

10. A method for diagnosing fetal chromosomal aneuploidy of a specified chromosome or chromosome region comprising

a) providing a sample comprising circulating cell-free DNA (cf-DNA) obtained from human maternal plasma,

b) partitioning the sample into a plurality of reaction volumes,

c) assaying each reaction volume for the presence of each of two or more DNA target sequences, wherein said target sequences are closely linked in the specified chromosome or chromosome region,

d) determining the number of reaction volumes that contain at least one DNA molecule comprising at least one of said two or more of DNA target sequences, e) comparing said number of reaction volumes with a reference number characteristic of a known or assumed chromosome ploidy, wherein when said number of reaction volume is higher or lower than the reference number it is indicative of fetal chromosomal aneuploidy of the specified chromosome or chromosome region.

1 1 . The method of any of claims 1 -10 in which the sample comprising cf-DNA is enriched for DNAs smaller than 500 bp.

12. The method of any of claims 1 -10 in which at least 5 closely linked target sequences are assayed.

13. The method of any of claims 1-10 in which at least 10 closely linked target sequences are assayed.

14. The method of any of claims 1 -10 wherein each target sequence is less than 300 bases in length.

15. The method of any of claims 1 -10 wherein each target sequence is less than 100 bases in length.

16. The method of any of claims 1 -10 in which the closely linked target sequences all are within 2 kb of each other on the chromosome.

17. The method of any of claims 1 -10 in which at least 10 target sequences are assayed and each target sequence is closely linked to at least one of the other target sequences.

18. The method of claim 17 wherein each target sequence is located in the chromosome within 500 bp of at least one of the other target sequences.

19. The method of claim 3 in which the closely linked target sequences are on chromosome 18 or 21 .

20. The method of claim 4 wherein the closely linked target sequences are in the ATM (ataxia telangiectasia mutated) locus.

21 . A method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

a) distributing a DNA sample comprising fragmented chromosomal DNA heterogeneous in size into a plurality of reaction volumes, wherein after said distribution on average each reaction volume contains less than one DNA fragment containing a first specified chromosomal target sequence, wherein said sample comprises a first population of DNA fragments in a first size range and a second population of DNA fragments in a second size range, and the mean of the second size range is greater than the mean of the first size range; b) assaying each reaction volume for the presence of at least one fragment containing said first specified chromosomal target sequence, wherein if such a fragment is present a first detectable signal is produced, and assaying each reaction volume for the presence of a fragment containing a second specified chromosomal target sequence, wherein if such a fragment is present a second detectable signal is produced,

wherein said first and second specified chromosomal target sequences are closely linked on the same chromosome and wherein said first and second

detectable signals are distinguishable;

c) determining from the number or proportion of reaction volumes in which the first detectable signal is produced and no second detectable signal is produced the presence or representation of fragments in the DNA sample that (i) comprise the first specified chromosomal target sequence and (ii) have a length in the first size range,

and/or

22. A method of differentially detecting one or more target sequences in a sample comprising a mixture of short and long polynucleotides, the method comprising: distributing the sample into discrete reaction volumes, each volume

comprising an average of no more than about one detectable polynucleotide per reaction volume;

distributing in each reaction volume at least a first detectable probe and a second detectable probe wherein the first and second probes are selected such that both the first and second probes will bind to a long polynucleotide and only the first or the second probe will bind to a short polynucleotide; detecting first and/or second probe binding to polynucleotide in a reaction volume, wherein detection of either the first or second probe binding indicates detection of probe binding to a short polynucleotide and detection of both first and second probe binding indicates detection of probe binding to a long polynucleotide, wherein a "detectable polynucleotide" is a polynucleotide that can be detected with the first probe and/or the second probe.

23. The method of claim 22 wherein the probes detect amplicons generated in the reaction volumes by amplification of sequences from said short and long

polynucleotides.

24. A method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

wherein in a formula describing a relationship of assays

"-A-" indicates that a target sequence in the first set is assayed;

25. The method of claim 24 further comprising

optionally assaying each reaction volume for the presence of one or more of a set of third target sequences, wherein if a fragment comprising one or more of said third target sequences is present in the reaction volume a signal C is detected;

"-C-" indicates that a target sequence in the third set is assayed;

each reaction volume is assayed for target sequences having one or more relationships selected from:

iii) A-(-B-)_N-A

iv) (-A-B-C-)

v) (-A-B-B-)N

wherein N = 1 -10 and each target sequence is independently selected; distinguishing reaction volumes in which a single signal is detected from reaction volumes in which more than one signal is detected, wherein a reaction volume in which a single signal is detected is counted as containing a fragment that comprises a target sequence and is from said first population and a reaction volume in more than one signal is detected is counted as containing a fragment that comprises a target sequence and is from said second population.

26. A method for genetic analysis comprising:

a) carrying out the method of claim 24 in which the target sequences are on a first chromosome

b) carrying out the method of claim 25 in which the target sequences are on a second chromosome, with the modification in assaying target sequences on the second chromosome signal A or signal B or both is distinguishable from signal A or signal B or both, respectively, on the first chromosome

27. A method for genetic analysis of heterogeneously sized chromosomal DNA fragments, said method comprising:

c) assaying each reaction volume for the presence of a second target sequence, wherein if a fragment comprising one or more of said second target sequences is present in the reaction volume a signal A' (A prime) is detected wherein the first and second target sequences are allelic variants in the same gene

wherein in a formula describing a relationship of assays

"-A-" indicates that a first target sequence is assayed;

"-A -" indicates that a second target sequence is assayed;

"-B-" indicates that a third target sequence is assayed;

"-C-" indicates that a fourth target sequence is assayed; and

i) B-A,

ii) A-B,

iii) B-A-C,

iv) B-A-B,

and one or more relationships selected from:

v) B-A',

vi) A'-B;

vii) B-A'-C

viii) B-A'-B

ix) C-A'-C

identifying reaction volumes in which only a signal selected from A is detected as comprising a fragment from the first population, said fragment comprising the first target sequence; identifying reaction volumes in which only a signal selected from A' is detected as comprising a fragment from the first population, said fragment

comprising the second target sequence.

28. The method of any one of claims 21 , 24, 26 or 27 wherein target sequences are closely linked on the chromosome if they are separated by less than 1 kbp.

29. The method of any one of claims 21 , 24, 26 or 27 wherein the DNA sample is circulating cell-free DNA from a human.

30. The method of claim 30 wherein the human is a pregnant woman.

31 . The method of any one of claims 21 , 24, 26 or 27 wherein DNA fragments in the first size range are enriched in fetal cf-DNA compared to total cf-DNA and DNA fragments in a second size range are enriched in maternal cf-DNA compared to total cf-DNA.

32. The method of any one of claims 21 , 24, 26 or 27 wherein the first size range is 25-200 bp and the second size range is 400 bp to 20 kbp.

33. The method of claim 30 wherein the human is a person diagnosed with or suspected of having, cancer.

34. The method of claim 21 wherein the first and second detectable signals are fluorescence.

35. The method of claim 21 wherein the first and second detectable signals are melting temperature.

36. The method of claim 21 wherein the combined lengths of the first specified chromosomal sequence, the second specified chromosomal sequence, and the distance between said sequences is more than 100 bp and less than 1000 bp.