Abstract
As we enter the age of personalized medicine, healthcare is increasingly focused on tailoring diagnoses and treatments based on patients' genetic and environmental circumstances. A critical component of a person's physiological makeup is their immune system, but individual genetic variation in many immune system genes has remained resistant to analysis using classical whole-genome or targeted sequencing approaches. In particular, germline adaptive immune system genes, like immunoglobulin (IG) and T cell receptor (TR) genes, are particularly hard to genotype using classic reference-based methods owing to their highly repetitive and homologous nature. In this paper, we present ImmunoTyper2, a new computational toolkit for genotyping the variable genes of the IG lambda and kappa, and the TR loci with short-read whole genome sequence data, using an integer linear programming formulation, as an update to the ImmunoTyper-SR suite, which focused on IGHV region only. We evaluate its genotyping performance using Mendelian concordance analysis in 590 trios from the 1000 Genomes Project, benchmarking 40 samples against HPRC assembly-derived genotypes, and assessing robustness through sequencing depth analysis and parameter sensitivity tests. We introduce allele call confidence metrics to help quantify reliability. We also perform a prospective disease association study, applying ImmunoTyper2 to a WGS data set from a cohort of 461 COVID-19 patients from the COVNET Consortium to demonstrate how it can be applied to investigate genetic associations with disease.