Sensitive and specific post-call filtering of genetic variants in xenograft and primary tumors.

Abstract

Tumor genome sequencing offers great promise for guiding research and therapy, but spurious variant calls can arise from multiple sources. Mouse contamination can generate many spurious calls when sequencing patient-derived xenografts. Paralogous genome sequences can also generate spurious calls when sequencing any tumor. We developed a BLAST-based algorithm, Mouse And Paralog EXterminator (MAPEX), to identify and filter out spurious calls from both these sources. When calling variants from xenografts, MAPEX has similar sensitivity and specificity to more complex algorithms. When applied to any tumor, MAPEX also automatically flags calls that potentially arise from paralogous sequences. Our implementation, mapexr, runs quickly and easily on a desktop computer. MAPEX is thus a useful addition to almost any pipeline for calling genetic variants in tumors. The mapexr package for R is available at https://github.com/bmannakee/mapexr under the MIT license. mannakee@email.arizona.edu or rgutenk@email.arizona.edu or eknudsen@email.arizona.edu. Supplementary data are available at Bioinformatics online.