Methods and systems for dynamic variant thresholding in a liquid biopsy assay

Abstract

Methods, systems, and software are provided for validating a somatic sequence variant in a subject having a cancer condition. Sequence reads are obtained from sequencing cell-free DNA fragments in a liquid biopsy sample of the subject. Sequence reads are aligned to a reference sequence. A variant allele fragment count and locus fragment count are identified for a candidate variant that maps to a locus in the reference sequence. The variant allele fragment count is compared against a dynamic variant count threshold for the locus. The threshold is based on a pre-test odds of a positive variant call for the locus, based on the prevalence of variants in a genomic region including the locus in a cohort of subjects having the cancer condition. The somatic sequence variant in the subject is validated, or rejected, when the variant allele fragment count for the candidate variant satisfies, or does not satisfy, the threshold.

What is claimed is: 
     
         1 . A method of validating a somatic sequence variant in a cancerous tissue of a test subject having a cancer condition, the method comprising:
 at a computer system having one or more processors, and memory storing one or more programs for execution by the one or more processors:   (A) obtaining, from a first sequencing reaction, a corresponding sequence of each cell-free DNA fragment in a first plurality of cell-free DNA fragments in a liquid biopsy sample of the test subject, thereby obtaining a first plurality of sequence reads, wherein the first plurality of sequence reads comprises at least 100,000 sequence reads;   (B) aligning each respective sequence read in the first plurality of sequence reads to a reference sequence for the species of the subject thereby identifying a candidate somatic sequence variant mapping to a respective locus in the reference sequence;   (C) determining for the candidate somatic sequence variant, (i) a respective variant allele fragment count for the first sequencing reaction, and (ii) a respective locus fragment count for the first sequencing reaction; and   (D) comparing the respective variant allele fragment count for the candidate somatic sequence variant against a dynamic variant count threshold for the respective locus in the reference sequence that the candidate variant maps to, wherein the dynamic variant count threshold is based upon at least a pre-test odds of a positive variant call for the respective locus based upon a prevalence of variants in a genomic region that includes the respective locus in a cohort of training subjects having the cancer condition, and:
 when the variant allele fragment count for the candidate somatic sequence variant satisfies the dynamic variant count threshold for the respective locus, not rejecting the presence of the candidate somatic sequence variant in the test subject, or 
 when the variant allele fragment count for the candidate somatic sequence variant does not satisfy the dynamic variant count threshold for the locus, rejecting the presence of the candidate somatic sequence variant in the test subject. 
   
     
     
         2 . A method of validating a somatic sequence variant in a test subject having a cancer condition, the method comprising:
 at a computer system having one or more processors, and memory storing one or more programs for execution by the one or more processors:   (A) obtaining, from a first sequencing reaction, a corresponding sequence of each cell-free DNA fragment in a first plurality of cell-free DNA fragments in a liquid biopsy sample of the test subject, thereby obtaining a first plurality of sequence reads;   (B) aligning each respective sequence read in the first plurality of sequence reads to a reference sequence for the species of the subject thereby identifying a candidate somatic sequence variant mapping to a respective locus in the reference sequence, wherein the reference sequence for the species represents at least 1 Mb of the genome for the species;   (C) determining for the candidate somatic sequence variant, (i) a respective variant allele fragment count for the first sequencing reaction, and (ii) a respective locus fragment count for the first sequencing reaction; and   (D) comparing the respective variant allele fragment count for the candidate somatic sequence variant against a dynamic variant count threshold for the respective locus in the reference sequence that the candidate variant maps to, wherein the dynamic variant count threshold is based upon at least a pre-test odds of a positive variant call for the respective locus based upon a prevalence of variants in a genomic region that includes the respective locus in a cohort of training subjects having the cancer condition, and:
 when the variant allele fragment count for the candidate somatic sequence variant satisfies the dynamic variant count threshold for the respective locus, not rejecting the presence of the candidate somatic sequence variant in the test subject, or 
 when the variant allele fragment count for the candidate somatic sequence variant does not satisfy the dynamic variant count threshold for the locus, rejecting the presence of the candidate somatic sequence variant in the test subject. 
   
     
     
         3 . The method of  claim 1  or  2 , wherein the dynamic variant count threshold is also based upon a sequencing error rate for the sequencing reaction. 
     
     
         4 . The method of  claim 3 , wherein the sequencing error rate for the sequencing reaction is a trinucleotide sequencing error rate. 
     
     
         5 . The method of any one of  claims 1 - 4 , wherein the dynamic variant count threshold is also based upon a background sequencing error rate determined for the locus. 
     
     
         6 . The method of any one of  claims 1 - 5 , method further comprising:
 determining the dynamic variant count threshold based upon a variant detection specificity determined according to the relationship:   
       
         
           
             
               specificity 
               = 
               
                 1 
                 - 
                 
                   ( 
                   
                     
                       ( 
                       sensitivity 
                       ) 
                     
                     × 
                     
                       ( 
                       
                         
                           odds 
                           ( 
                           
                             pre 
                             - 
                             test 
                           
                           ) 
                         
                         
                           odds 
                           ( 
                           
                             post 
                             - 
                             test 
                           
                           ) 
                         
                       
                       ) 
                     
                   
                   ) 
                 
               
             
           
         
         wherein,
 sensitivity is a variant detection sensitivity selected from a distribution of variant detection sensitivities based on an estimated circulating variant fraction for the candidate variant, 
 odds(post-test) is a post-test odds of a positive variant call for the locus, and 
 odds(pre-test) is the pre-test odds of the positive variant call for the locus. 
 
       
     
     
         7 . The method of  claim 6 , wherein the distribution of variant detection sensitivities is based on a correlation between (i) the detection rate of a reference variant allele, in one or more sequencing reactions that are process-matched with the first sequencing reaction, for a plurality of cancer samples, and (ii) the variant allele fractions for the reference variant allele in the cancer samples. 
     
     
         8 . The method of  claim 7 , wherein the correlation is established by determining, for each respective bin in a plurality of bins collectively representing a span of variant allele fractions represented in the cancer samples, wherein each respective bin corresponds to a contiguous span of variant allele fractions that does not overlap with any other respective bin in the plurality of bins, a corresponding sensitivity for detection of the reference variant alleles for the corresponding subset of cancer samples. 
     
     
         9 . The method of any one of  claims 6 - 8 , wherein the estimated circulating variant fraction for the candidate variant is a variant allele fraction determined from a comparison of (i) the respective variant allele fragment count for the first sequencing reaction, to (ii) the respective locus fragment count for the first sequencing reaction 
     
     
         10 . The method of any one of  claims 6 - 9 , wherein the specificity is used to select a quantile of a beta-binomial distribution of the minimal variant allele fragment count required to support a positive variant call for the locus, thereby defining the dynamic threshold for the locus, wherein the beta-binomial distribution is defined by a sequencing error rate for the sequencing reaction and a background sequencing error rate determined for the locus. 
     
     
         11 . The method of any one of  claims 6 - 10 , wherein the pre-test odds of a positive variant call for the locus is based on the prevalence of variants in a genomic region that includes the locus from the first set of nucleic acids obtained from the cohort of subjects having the cancer condition. 
     
     
         12 . The method of  claim 11 , wherein, when the genomic region that includes the locus is associated with a mutation known to confer resistance against a therapy used to treat the cancer condition, the pre-test odds are boosted based on a pre-test-odds multiplier specific for the genomic region. 
     
     
         13 . The method of  claim 11  or  12 , wherein the pre-test odds of a positive variant call for the locus is further based on a known or inferred effect of the variants, wherein:
 when the known or inferred effect of a variant is loss-of-function of a gene that includes the locus, the genomic region used to compute the pre-test probability is the entire gene, and 
 when the known or inferred effect of a variant is gain-of-function of the gene that includes the locus, the genomic region used to compute the pre-test probability is the exon, of the gene, that includes the locus. 
 
     
     
         14 . The method of  claim 13 , wherein the effect of the variants is inferred by:
 binning each respective variant of the variants in the genomic region that includes the locus from the first set of nucleic acids obtained from the cohort of subjects having the cancer condition into a respective bin, in a plurality of bins for the gene that include the locus, corresponding to the exon encompassing the respective variant in the gene, wherein each bin in the plurality of bins corresponds to a different exon of the respective gene; and   determining whether any bin in the plurality of bins contains significantly more variants than the other bins in the plurality of bins, wherein:
 when a bin contains significantly more variants than the other bins in the plurality of bins, the effect of the sequence variant is inferred to be a gain-of-function of the gene, and 
 when no bin in the plurality of bins contains significantly more sequence variants than the other bins in the plurality of bins, the effect of the sequence variant is inferred to be a loss-of-function of the gene. 
   
     
     
         15 . The method of  claim 14 , wherein determining whether any bin in the plurality of bins contains significantly more variants than the other bins in the plurality of bins comprises applying a rolling Poisson test of difference between bin counts corresponding to adjacent exons in the gene. 
     
     
         16 . The method of any one of  claims 1 - 15 , wherein the liquid biopsy sample is blood. 
     
     
         17 . The method of any one of  claims 1 - 15 , wherein the liquid biopsy sample comprises blood, whole blood, peripheral blood, plasma, serum, or lymph of the test subject. 
     
     
         18 . The method of any one of  claims 1 - 17 , wherein:
 the first sequencing reaction is a panel-enriched sequencing reaction of a first plurality of enriched loci, and   each respective locus in the plurality of enriched loci are sequenced at an average unique sequence depth of at least 250×.   
     
     
         19 . The method of  claim 18 , wherein each respective locus in the plurality of enriched loci are sequenced at an average unique sequence depth of at least 1000×. 
     
     
         20 . The method of  claim 18  or  19 , wherein the panel-enriched sequencing reaction uses a sequencing panel that enriches for at least 50 genes. 
     
     
         21 . The method of any one of  claims 18 - 20 , wherein the panel-enriched sequencing reaction uses a sequencing panel that enriches for at least 10 genes listed in Table 1. 
     
     
         22 . The method of any one of  claims 18 - 21 , wherein the panel-enriched sequencing reaction uses a sequencing panel that enriches for at least 10 genes listed in List 1. 
     
     
         23 . The method of any one of  claims 18 - 22 , wherein the panel-enriched sequencing reaction uses a sequencing panel that enriches for at least 10 genes listed in List 2. 
     
     
         24 . The method of any one of  claims 1 - 23 , wherein:
 the first sequencing reaction is a whole genome sequencing reaction, and   the average sequencing depth of the reaction across the genome is at least 25×.   
     
     
         25 . The method of any one of  claims 1 - 24 , wherein the first plurality of sequence reads comprises at least 50,000 sequence reads. 
     
     
         26 . The method of any one of  claims 1 - 24 , wherein the first plurality of sequence reads comprises at least 250,000 sequence reads. 
     
     
         27 . The method of any one of  claims 1 - 26 , wherein the cancer condition is a particular type and stage of cancer. 
     
     
         28 . The method of any one of  claims 1 - 27 , wherein the cohort of subjects having the cancer condition are matched to at least one personal characteristic of the subject. 
     
     
         29 . The method of any one of  claims 1 - 27 , wherein, when the variant allele fragment count for the candidate variant satisfies the dynamic variant count threshold for the locus and all other variant calling 
     
     
         30 . The method of any one of  claims 1 - 28 , further comprising generating a report for the test subject comprising the identity of variant alleles having variant allele counts, in the first sequencing reaction, that satisfy the dynamic variant count threshold. 
     
     
         31 . The method of  claim 30 , wherein the generated report further comprises therapeutic recommendations for the test subject based on the identity of one or more of the reported variant alleles. 
     
     
         32 . A computer system comprising:
 one or more processors; and   a non-transitory computer-readable medium including computer-executable instructions that, when executed by the one or more processors, cause the processors to perform a method according to any one of  claims 1 - 31 .   
     
     
         33 . A non-transitory computer-readable storage medium having stored thereon program code instructions that, when executed by a processor, cause the processor to perform the method according to any one of  claims 1 - 31 .