Deep learning analysis pipeline for next generation sequencing
A method for variant calling in a next generation sequencing analysis pipeline involves obtaining a plurality of sequence reads that each include a nucleotide aligned at a nucleotide position within a sample genome. The method also involves obtaining a plurality of alleles associated with the nucleotide position. The method further involves determining that a particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, wherein the particular allele is located at the nucleotide position. Additionally, the method involves generating an image based on information associated with the plurality of sequence reads. Further, the method involves determining, by providing the generated image to a trained neural network, a likelihood that the sample genome contains the particular allele. The method may also involves providing an output signal indicative of the determined likelihood.
1. A method comprising:
obtaining a plurality of sequence reads that each include a nucleotide aligned at a nucleotide position within a sample genome;
obtaining a plurality of alleles associated with the nucleotide position;
determining that a particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, wherein the particular allele is located at the nucleotide position;
generating an image based on information associated with the plurality of sequence reads;
determining, by providing the generated image to a trained neural network, a likelihood that the sample genome contains the particular allele; and
providing an output signal indicative of the determined likelihood.
2. The method of claim 1 , wherein the trained neural network is a convolutional neural network.
3. The method of claim 1 , wherein the plurality of alleles are data objects specifying one or more of the following: (i) a position in the sample genome, (ii) a label, and (iii) a type of variant.
4. The method of 3 , further comprising:
upon determining that the particular allele of the plurality of alleles matches one or more sequence reads, appending the number of matching sequence reads to the data object associated with the particular allele.
5. The method of 3 , further comprising:
upon determining that the particular allele of the plurality of alleles matches one or more sequence reads, appending a number of disambiguating sequence reads that indicate a presence of unknown alleles at the nucleotide position to the data object associated with the particular allele.
6. The method of claim 1 , wherein generating the image comprises:
generating color information for a plurality of pixels that collectively form the image, wherein the color information includes a plurality of color channels, wherein a first color channel represents a nucleotide base type, a second color channel represents a quality score, and a third color channel represents a presence of a nucleotide base.
7. The method of claim 1 , wherein the output signal indicative of the determined likelihood includes data stored in a variant call format (VCF).
8. The method of claim 1 , wherein the plurality of sequence reads is a first plurality of sequence reads, and wherein the method further comprises:
performing, for a second plurality of sequence reads that each include a nucleotide aligned at a second nucleotide position within the sample genome, the operations of (i) obtaining the plurality of sequence reads, (ii) obtaining the plurality of alleles, (iii) determining that the particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, (iv) generating the image, (v) determining the genotype likelihoods and probability of being an error, and (vi) providing the output signal in parallel with the operations for the first plurality of sequence reads.
9. A method comprising:
obtaining a plurality of sequence reads that each include a nucleotide aligned at a nucleotide position within a sample genome, wherein the sequence reads are aligned to a reference genome according to a first alignment;
receiving a variant call at the nucleotide position and associated with the sample genome, wherein the variant call includes a first confidence value indicative of a likelihood that the sample genome contains a variant specified by the variant call at the nucleotide position;
determining that the nucleotide position is located within a region of ambiguity;
responsive to receiving the variant call at the nucleotide position and determining that the nucleotide position is located within the region of ambiguity, realigning one or more sequence reads of the plurality of sequence reads, wherein the realignment causes the sequence reads to be aligned to the reference genome according to a second alignment; and
providing an output signal indicative of the second alignment.
10. The method of claim 9 , wherein determining that the nucleotide position is located within a region of ambiguity comprises:
obtaining a set of variant calls at positions adjacent to the nucleotide position that include a respective set of confidence values, wherein the positions of the set of variant calls define the region of ambiguity;
determining that the average of the set of confidence values is below a threshold level of confidence; and
based on the determination that the average of the set of confidence values is below the threshold level of confidence, labeling the nucleotide position as being within the region of ambiguity, wherein labeling the nucleotide position as being within the region of ambiguity causes execution of the realignment.
11. The method of claim 9 , wherein realigning the one or more sequence reads comprises:
performing a local string graph assembly procedure on the plurality of sequence reads to determine one or more potential haplotypes; and
realigning the plurality of sequence reads based on the one or more potential haplotypes.
12. The method of claim 11 , wherein performing the local string graph assembly procedure comprises providing the plurality of sequence reads to a De Bruijn graph.
13. The method of claim 9 , further comprising:
based on the second alignment, determining a second confidence value indicative of a likelihood that the sample genome contains the variant specified by the variant call at the nucleotide position after realignment, wherein the output signal includes the second confidence value.
14. A non-transitory computer-readable medium having instructions stored thereon that, upon execution by at least one processor, causes performance of operations comprising:
obtaining a plurality of sequence reads that each include a nucleotide aligned at a nucleotide position within a sample genome;
obtaining a plurality of alleles associated with the nucleotide position;
determining that a particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, wherein the particular allele is located at the nucleotide position;
generating an image based on information associated with the plurality of sequence reads;
determining, by providing the generated image to a trained neural network, a likelihood that the sample genome contains the particular allele; and
providing an output signal indicative of the determined likelihood.
15. The non-transitory computer-readable medium of claim 14 , wherein the trained neural network is a convolutional neural network.
16. The non-transitory computer-readable medium of claim 14 , wherein the plurality of alleles are data objects specifying one or more of the following: (i) a position in the sample genome, (ii) a label, and (iii) a type of variant.
17. The non-transitory computer-readable medium of claim 16 , further comprising:
upon determining that the particular allele of the plurality of alleles matches one or more sequence reads, appending the number of matching sequence reads to the data object associated with the particular allele.
18. The non-transitory computer-readable medium of claim 16 , further comprising:
upon determining that the particular allele of the plurality of alleles matches one or more sequence reads, appending a number of disambiguating sequence reads that indicate a presence of unknown alleles at the nucleotide position to the data object associated with the particular allele.
19. The non-transitory computer-readable medium of claim 14 , wherein the output signal indicative of the determined likelihood includes data stored in a variant call format (VCF).
20. The non-transitory computer-readable medium of claim 14 , wherein the plurality of sequence reads is a first plurality of sequence reads, and wherein the method further comprises:
performing, for a second plurality of sequence reads that each include a nucleotide aligned at a second nucleotide position within the sample genome, the operations of (i) obtaining the plurality of sequence reads, (ii) obtaining the plurality of alleles, (iii) determining that the particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, (iv) generating the image, (v) determining the likelihood, and (vi) providing the output signal in parallel with the operations for the first plurality of sequence reads.