Methods and applications of gene fusion detection in cell-free DNA analysis

A sequencing system for identifying first and second breakpoints of a fusion gene associated with a disease or condition, comprising:

• a communication interface configured to receive, over a communication network, genetic sequence read data from a sample of nucleic acid molecules from a subject generated by a nucleic acid sequencer, the genetic sequence read data comprising a paired-end sequencing reads 50 to 150 bases in length of at least 1 million polynucleotide molecules;
• and
• a computer in communication with the communication interface, the computer comprising one or more processors and a non-transitory machine-readable medium comprising instructions that, upon execution by the one or more processors, cause the computer to at least:
• merge pairs of pair-end sequencing reads having a computed insert size less than a sum of lengths of the paired-end sequencing reads, thereby producing merged reads that minimize redundant alignment operations;
• store the merged reads in a breakpoint read list data structure that retains a unique read identifier and clipping information for each merged read;
• search a whole human reference genome for alignments of the merged reads wherein the search is configured to detect and preserve partial alignments indicating clipped portions, thereby producing mapped merged reads;
• extract from the mapped merged reads, for each mapper merge read, one or more clipped portions, wherein a mapped read comprises:
• a first clipped portion that aligns to a first region of the whole human reference genome and
• a second clipped portion that aligns to a second region of the whole human reference genome distinct from the first region,
• is identified as a fused read;
• determine a respective candidate breakpoint pair for each fused read based on genomic positions at which the fused read is clipped, wherein the candidate breakpoint pair is stored in the breakpoint read list data structure and is linked to the unique read identifier;
• determine a plurality of sets of fused reads, wherein each set of fused reads of the plurality of sets of fused reads comprises a plurality of the fused reads having a common candidate breakpoint pair;
• generate one or more clusters of sets of fused reads, individual clusters of the one or more clusters including one or more sets of fused reads having candidate breakpoints within a predetermined genetic distance of one another and, for each cluster of the one or more clusters,
• (i) store a count of wild-type coverage crossing each candidate breakpoint,
• (ii) store a count of fused reads supporting each candidate breakpoint, and/or
• (iii) compute a metric comparing fused coverage to wild-type coverage;
• select a final first breakpoint and a final second breakpoint for the fusion gene by applying at least one selection criterion to the candidate breakpoints in each cluster of the one or more of the clusters, wherein the at least one selection criterion requires at least one of:
• (i) a minimum number of unique read identifiers and an observed reduction in false-positive alignments, or
• (ii) that the metric comparing fused coverage to wild-type coverage exceeds a threshold value,
• thereby reducing false-positive alignments and enabling increased specificity in fusion detection;
• and
• determine based on the final first breakpoint and the final second breakpoint of the fusion gene at least one of:
• at least one of a diagnosis, a prognosis, or a risk of developing the disease or condition in the subject,
• a guide for treatment of the disease or condition, or
• a guide to preparation of a medicament for treatment of the disease or condition.