Detecting fetal sub-chromosomal aneuploidies
Disclosed are methods for determining copy number variation (CNV) known or suspected to be associated with a variety of medical conditions, including syndromes related to CNV of subchromosomal regions. In some embodiments, methods are provided for determining CNV of fetuses using maternal samples comprising maternal and fetal cell free DNA. Some embodiments disclosed herein provide methods to improve the sensitivity and/or specificity of sequence data analysis by removing within-sample GC-content bias. In some embodiments, removal of within-sample GC-content bias is based on sequence data corrected for systematic variation common across unaffected training samples. In some embodiments, syndrome related biases in sample data are also removed to increase signal to noise ratio. Also disclosed are systems for evaluation of CNV of sequences of interest.
1. A method, implemented at a computer system that includes one or more processors and system memory, for evaluation of copy number of a sequence of interest in a test sample comprising nucleic acids, the method comprising:
(a) sequencing DNA in the test sample using a sequencer to obtain sequence reads;
(b) obtaining, by the computer system, the sequence reads;
(c) aligning, by the computer system, the sequence reads of the test sample to a reference genome comprising the sequence of interest, thereby providing test sequence tags, wherein the reference genome is divided into a plurality of bins, wherein the sequence of interest is in a sub-chromosomal genomic region in which a copy number variation is associated with a genetic syndrome;
(d) determining, by the computer system, coverages of the test sequence tags for bins in the sequence of interest;
(e) adjusting, by the computer system, the coverages of the test sequence tags for the bins in the sequence of interest using expected coverages for the bins in the sequence of interest obtained from a subset of a training set of unaffected training samples without using unaffected samples outside the subset, wherein training samples in the subset were obtained by:
selecting unaffected samples from the training set of unaffected training samples that are more highly correlated with each other than training samples of the training set of unaffected training samples not in the subset are correlated with each other, and wherein the correlation is based on coverages in a plurality of bins outside the sequence of interest; and
(f) making, by the computer system, a call of the copy number variation of the sequence of interest in the test sample based on the adjusted coverages from (e).
2. The method of claim 1 , wherein the plurality of bins outside the sequence of interest are more highly correlated with at least one bin in the sequence of interest than one or more bins outside the sequence of interest are correlated with the at least one bin in the sequence of interest.
3. The method of claim 2 , wherein the plurality of bins outside the sequence of interest are identified by determining correlation distances between a coverage in a bin under consideration within the sequence of interest and a coverage of each individual bin of the bins outside the sequence of interest.
4. The method of claim 3 , wherein the correlation distances are calculated as distances between vectors of bin coverages created from samples of the training set.
5. The method of claim 1 , further comprising determining, based on (f), whether one or more of genomes in the test sample has a chromosomal aneuploidy.
6. The method of claim 1 , wherein the bins outside the sequence of interest used in (e) are bins in one or more human autosomes other than chromosomes 13, 18, and 21.
7. The method of claim 1 , wherein the subset of the training set is selected by identifying a cluster of the training samples in the training set.
8. The method of claim 1 , further comprising repeating (e) for a number of iterations, wherein each iteration uses adjusted coverages from a previous iteration as the coverages to be adjusted in a current iteration, and wherein each iteration employs expected coverages obtained from a different subset of the training set.
9. The method of claim 1 , wherein adjusting the coverages of the test sequence tags for the bins in the reference genome in operation (e) comprises:
fitting a function to data points, each data point relating an expected coverage to a corresponding coverage for the test sample in a bin; and
adjusting the coverages for the bins by applying the coverages for the bins to the function.
10. The method of claim 9 , wherein the function is a linear function.
11. The method of claim 1 , further comprising performing segmentation to determine start and end points of a syndrome specific region as the sequence of interest.
12. The method of claim 1 , wherein the test sample comprises a mixture of nucleic acids from two different genomes.
13. The method of claim 1 , wherein said DNA comprises cfDNA molecules.
14. The method of claim 1 , wherein the test sample comprises fetal and maternal cell-free nucleic acids.
15. The method of claim 1 , wherein the test sample comprises nucleic acids from cancerous and unaffected cells from the same subject.
16. The method of claim 1 , wherein the sequencing reads are obtained by an initial multiplex sequencing, further comprising:
determining that the test sample has a first value for calling a syndrome classification or a copy number variation higher than a first threshold;
resequencing the test sample at a sequencing depth deeper than the initial multiplex sequencing to obtain resequenced data; and
determining the syndrome classification or the copy number variation using the resequenced data.
17. The method of claim 1 , wherein the genetic syndrome is selected from the group consisting of: 1p36 deletion syndrome, Wolf-Hirschhorn syndrome, Cri-du-Chat syndrome, Angelman syndrome, Williams syndrome, and DiGeorge syndrome, and the method further comprises diagnosing the genetic syndrome.
18. The method of claim 1 , wherein the training samples in the subset of the training set of unaffected training samples have a common systematic variability in normalized coverage.
19. The method of claim 1 , further comprising reporting, based on the call of the copy number variation, a diagnosis of a genetic syndrome.
20. A computer program product comprising a non-transitory machine readable medium storing program code that, when executed by one or more processors of a computer system, causes the computer system to implement a method for evaluation of copy number of a sequence of interest, said program code comprising code for:
(a) causing a sequencer to sequence DNA in the test sample to obtain sequence reads;
(b) obtaining the sequence reads;
(c) aligning the sequence reads of the test sample to a reference genome comprising the sequence of interest, thereby providing test sequence tags, wherein the reference genome is divided into a plurality of bins, wherein the sequence of interest is in a sub-chromosomal genomic region in which a copy number variation is associated with a genetic syndrome;
(d) determining coverages of the test sequence tags for bins in the sequence of interest;
(e) adjusting the coverages of the test sequence tags for the bins in the sequence of interest using expected coverages for the bins in the sequence of interest obtained from a subset of a training set of unaffected training samples without using unaffected samples outside the subset, wherein training samples in the subset were obtained by:
selecting unaffected samples from the training set of unaffected training samples that are more highly correlated with each other than training samples of the training set of unaffected training samples not in the subset are correlated with each other, and wherein the correlation is based on coverages in a plurality of bins outside of the sequence of interest; and
(f) making a call of the copy number variation of the sequence of interest in the test sample based on the adjusted coverages from (e).
21. A system for evaluation of copy number of a sequence of interest related to a genetic syndrome using a test sample comprising nucleic acids, the system comprising: a sequencer for receiving nucleic acids from the test sample and providing nucleic acid sequence information from the test sample; and logic designed or configured to execute or cause the following operations: (a) sequencing DNA in the test sample using the sequencer to obtain sequence reads; (b) obtaining the sequence reads; (c) aligning the sequence reads of the test sample to a reference genome comprising the sequence of interest, thereby providing test sequence tags, wherein the reference genome is divided into a plurality of bins, wherein the sequence of interest is in a sub-chromosomal genomic region in which a copy number variation is associated with the genetic syndrome; (d) determining coverages of the test sequence tags for bins in the sequence of interest; (e) adjusting the coverages of the test sequence tags for the bins in the sequence of interest using expected coverages for the bins in the sequence of interest obtained from a subset of a training set of unaffected training samples, wherein training samples in the subset were obtained by: selecting unaffected samples from the training set of unaffected training samples that are more highly correlated with each other than training samples of the training set of unaffected training samples not in the subset are correlated with each other, and wherein the correlation is based on coverages in a plurality of bins outside of the sequence of interest; and (f) making a call of the copy number variation of the sequence of interest in the test sample based on the adjusted coverages from (e).