Chromosomal and sub-chromosomal copy number variation detection

1 . A computer-implemented method for detecting a presence or absence of copy number variation in a target sample, comprising:

sequencing, by a massively parallel sequencer, nucleic acids from a plurality of samples as a batch run to generate sequencing data for the plurality of samples, wherein the plurality of samples includes the target sample, and wherein one or more of the plurality of samples are from different subjects;

obtaining, by a computing device, the sequencing data for the plurality of samples;

generating modified sequencing data for the plurality of samples by applying one or more mapping rules to the sequencing data, wherein the modified sequencing data comprises, for each sample of the plurality of samples: (i) a statistical coverage for each segment in the sample, and (ii) a global statistical coverage for all segments in the sample;

determining, by the computing device, for each sample of the plurality of samples, a first normalized coverage for each segment in the sample, wherein the determining the first normalized coverage for each segment in the sample comprises determining a segment statistical coverage ratio for the segment in the sample based on the statistical coverage for the segment and the global statistical coverage for all segments in the sample;

determining, by the computing device, for each sample of the plurality of samples, a second normalized coverage for each segment in the sample, wherein the determining the second normalized coverage for each segment in the sample comprises determining a copy number for the segment in the sample based on the first normalized coverage for the segment in the sample and a measure of central tendency of the first normalized coverage for the segment in all of the plurality of samples;

training, by the computing device, a machine learning model using a training set of training samples to classify the presence or absence of copy number variation, wherein the training set comprises a set of copy numbers corresponding to the training samples and corresponding ground truth labels for the presence or absence of copy number variation, wherein the training comprises:

inputting the set of copy numbers and the ground truth labels into the machine learning model,

adjusting parameters of the machine learning model to minimize a loss function indicative of classification error, and

validating the machine learning model using a validation set distinct from the training set;

inputting, by the computing device, the copy number for each segment in a target set in the target sample into the trained machine learning model, wherein the target set comprises one or more segments from all of the segments in the target sample; and

outputting, by the computing device, a classification for the presence or absence of the copy number variation for each segment in the target set in the target sample according to the classification by the trained machine learning model.

2 . The computer-implemented method of claim 1 , wherein the generating the modified sequencing data for the plurality of samples comprises:

mapping sequence reads in the sequencing data to a reference genome;

counting the sequence reads mapped to genomic portions of the reference genome, wherein the counting generates a quantification of the sequence reads mapped to the genomic portions of the reference genome for each sample;

determining a base coverage for each reference base within each segment of each sample based on the quantification of the sequence reads mapped to each reference base within each segment;

determining the statistical coverage for each segment in each sample; and

determining the global statistical coverage for all segments in each sample.

3 . The computer-implemented method of claim 1 , wherein the sequencing the nucleic acids generates hundreds of thousands to hundreds of millions of sequence reads for each sample.

4 . The computer-implemented method of claim 1 , wherein the statistical coverage for each segment in a sample is a median coverage for each segment in the sample, the global statistical coverage for all segments in the sample is a global median coverage for all segments in the sample, the measure of central tendency of the first normalized coverage for the segment in all of the plurality of samples is a mean of the first normalized coverage for the segment in all of the plurality of samples, and the determining the segment statistical coverage ratio for the segment in the sample comprises dividing the median coverage for the segment by the global median coverage for all segments in the sample.

5 . The computer-implemented method of claim 1 , wherein the determining the second normalized coverage for each segment in each of the samples further comprises removing outliers from the statistical coverage for the segment in each sample of the plurality of samples before determining the segment statistical coverage ratio for the segment in each sample of the plurality of samples, and calculating a mean of the segment statistical coverage ratio for the segment in remaining samples of the plurality of samples.

6 . The computer-implemented method of claim 1 , further comprising (i) removing one or more statistical coverage for one or more segments on a sex chromosome in one or more samples from the obtained sequencing data, (ii) removing one or more statistical coverage for one or more segments on Chromosome X in one or more samples associated with male subjects from the obtained sequencing data, (iii) doubling one or more statistical coverage for one or more segments on Chromosome X in one or more samples associated with male subjects in the obtained sequencing data, or (iv) any combination thereof.

7 . The computer-implemented method of claim 1 , further comprising:

determining, by the computing device, a first normalized coverage for an element in each sample of the plurality of samples, wherein the element is (i) a chromosome, (ii) a portion of the chromosome, (iii) a gene, (iv) an exon, (v) an intron, or (vi) a predetermined genomic part or region of interest, wherein the segment is a part of the element, and wherein the determining the first normalized coverage for the element in each sample comprises:

determining a local statistical coverage for the element in the sample; and

determining an element statistical coverage ratio for the element in the sample based on the local statistical coverage for the element and the global statistical coverage for all segments in the sample;

determining, by the computing device, a second normalized coverage for the element in each sample of the plurality of samples, the determining the second normalized coverage for the element in each sample comprises determining a copy number for the element in the sample based on the first normalized coverage for the element in the sample and a measure of central tendency of the first normalized coverage for the element in all of the plurality of samples;

inputting, by the computing device, the copy number for the element in the target sample into the trained machine learning model; and

outputting, by the computing device, a classification for the presence or absence of the copy number variation for the element in the target sample according to the classification by the trained machine learning model.

8 . The computer-implemented method of claim 7 , wherein the determining the second normalized coverage for the element further comprises removing outliers from the element statistical coverage ratio for the element in all of the plurality of samples before determining the measure of central tendency of the first normalized coverage for the element in all of the plurality of samples, and calculating a mean of the element statistical coverage ratio for the element in remaining samples of the plurality of samples.

9 . The computer-implemented method of claim 8 , wherein the removing the outliers comprises:

calculating an interquartile range based on the element statistical coverage ratio for the element in all of the plurality of samples;

determining an upper limit and a lower limit relating to the interquartile range; and

removing any element statistical coverage ratio for the element where a value of the element statistical coverage ratio is greater than the upper limit or less than the lower limit.

10 . The computer-implemented method of claim 1 , further comprising determining, by the computing device, whether the copy number for each segment in the target set in the target sample is within or outside of a predetermined interval based on the classification, wherein the outputting comprises reporting each segment in the target set in the target sample that has a copy number outside the predetermined interval as having the copy number variation, and reporting each segment in the target set in the target sample that has a copy number within the predetermined interval as being normal or not having the copy number variation.

11 . The computer-implemented method of claim 7 , further comprising determining, by the computing device, whether the copy number for the element in the target sample is within or outside of a predetermined interval based on the classification, wherein the outputting comprises reporting the element as having the copy number variation when the element in the target sample has a copy number outside the predetermined interval, or reporting the element as being normal or not having the copy number variation when the element in the target sample has a copy number within the predetermined interval.

12 . The computer-implemented method of claim 11 , further comprising:

calculating a statistical measure for each segment in the target set and/or the element, wherein the calculation of the statistical measure is based on a normal profile comprising the copy number for each segment in the target set and/or the element in all samples; and

comparing the statistical measure with a predetermined upper-threshold or with a predetermined lower-threshold set for segments and/or the element,

wherein each segment in the target in the target sample that has a copy number within the predetermined interval is only reported as having the copy number variation when the statistical measure of the segment in the target set is greater than the predetermined upper-threshold or less than the predetermined lower-threshold based on the comparing; and/or

wherein the element in the target sample that has a copy number within the predetermined interval is only reported as having the copy number variation when the statistical measure of the element is greater than the predetermined upper-threshold or less than the predetermined lower-threshold based on the comparing.

13 . The computer-implemented method of claim 12 , wherein the statistical measure comprises (i) a z-score, (ii) a p-value, or (iii) a coefficient of variation.

14 . The computer-implemented method of claim 1 , further comprising determining a diagnosis of a subject associated with the target sample, wherein the diagnosis is determined based on the classification for the presence or absence of the copy number variation for each segment in the target set in the target sample.

15 . The computer-implemented method of claim 14 , further comprising administering a treatment to the subject based on (i) the classification for the presence or absence of the copy number variation for each segment in the target set in the target sample, and/or (ii) the diagnosis of the subject.

16 . A computer-implemented method for detecting a presence or absence of copy number variation in a target sample, comprising:

sequencing, by a massively parallel sequencer, nucleic acids from a plurality of samples as a batch run to generate sequencing data for the plurality of samples wherein the plurality of samples includes the target sample, and wherein one or more of the samples are from different subjects;

obtaining, by a computing device, the sequencing data for the plurality of samples;

generating modified sequencing data for the plurality of samples by applying one or more mapping rules to the sequencing data, wherein the modified sequencing data comprises, for each sample of the plurality of samples: (i) a median coverage for each segment in the sample, and (ii) a global median coverage for all segments in the sample;

determining, by the computing device for each sample of the plurality of samples, a first normalized coverage for an element in the sample, wherein the element comprises two or more segments, and wherein the determining the first normalized coverage for the element comprises:

determining a local median coverage for the element in the sample; and

determining an element median coverage ratio for the element in the sample based on the local median coverage for the element and the global median coverage for all segments in the sample;

determining, by the computing device for each sample of the plurality of samples, a second normalized coverage for the element in the sample, wherein the determining the second normalized coverage for the element comprises determining a copy number for the element in the sample based on the first normalized coverage for the element in the sample and a mean of the first normalized coverage for the element in all of the plurality of samples;

training, by the computing device, a machine learning model using a training set of training samples to classify the presence or absence of copy number variation, wherein the training set comprises a set of copy numbers corresponding to the training samples and corresponding ground truth labels for the presence or absence of copy number variation, wherein the training comprises:

inputting the set of copy numbers and the ground truth labels into the machine learning model,

adjusting parameters of the machine learning model to minimize a loss function indicative of classification error, and

validating the machine learning model using a validation set distinct from the training set;

inputting, by the computing device, the copy number for the element in the target sample into the trained machine learning model; and

outputting, by the computing device, a classification for the presence or absence of the copy number variation for the element in the target sample according to the classification by the trained machine learning model.

17 . A system comprising:

a massively parallel sequencer, wherein the massively parallel sequencer sequences nucleic acids obtained from a plurality of samples as a batch run to generate sequencing data for the plurality of samples, wherein the plurality of samples includes a target sample, and wherein one or more of the plurality of samples are from different subjects;

a processor; and

a non-transitory memory comprising computer program instructions that when executed by the processor, cause the processor to perform

obtaining the sequencing data for the plurality of samples generated by the massively parallel sequencer;

generating modified sequencing data for the plurality of samples by applying one or more mapping rules to the sequencing data, wherein the modified sequencing data comprises, for each sample of the plurality of samples: (i) a median coverage for each segment in the sample, and (ii) a global median coverage for all segments in the sample;

determining, for each sample of the plurality of samples, a first normalized coverage for an element in the sample, wherein the element comprises two or more segments, and wherein the determining the first normalized coverage for the element comprises:

determining a local median coverage for the element in the sample; and

determining an element median coverage ratio for the element in the sample based on the local median coverage for the element and the global median coverage for all segments in the sample;

determining, for each sample of the plurality of samples, a second normalized coverage for the element in the sample, wherein the determining the second normalized coverage for the element comprises determining a copy number for the element in the sample based on the first normalized coverage for the element in the sample and a mean of the first normalized coverage for the element in all of the plurality of samples;

training a machine learning model using a training set of training samples to classify a presence or absence of copy number variation, wherein the training set comprises a set of copy numbers corresponding to the training samples and corresponding ground truth labels for the presence or absence of copy number variation, wherein the training comprises:

inputting the set of copy numbers and the ground truth labels into the machine learning model,

adjusting parameters of the machine learning model to minimize a loss function indicative of classification error, and

validating the machine learning model using a validation set distinct from the training set;

inputting the copy number for the element in the target sample into the trained machine learning model; and

outputting a classification for a presence or absence of a copy number variation for the element in the target sample according to the classification by the trained machine learning model.

18 . The computer-implemented method of claim 16 , further comprising determining, by the computing device, whether the copy number for the element in the target sample is within or outside of a predetermined interval based on the classification, wherein the outputting comprises reporting the element in the target sample that has a copy number outside the predetermined interval as having the copy number variation, and reporting the element in the target sample that has a copy number within the predetermined interval as being normal or not having the copy number variation.

19 . The system of claim 17 , wherein the processor is further caused to perform determining whether the copy number for the element in the target sample is within or outside of a predetermined interval based on the classification, wherein the outputting comprises reporting the element in the target sample that has a copy number outside the predetermined interval as having the copy number variation, and reporting the element in the target sample that has a copy number within the predetermined interval as being normal or not having the copy number variation.