Abstract

Analysis of antibody repertoires by high-throughput sequencing is of major importance in understanding adaptive immune responses. Our knowledge of variations in the genomic loci encoding antibody genes is incomplete, mostly due to technical difficulties in aligning short reads to these highly repetitive loci. The partial knowledge results in conflicting V-D-J gene assignments between different algorithms, and biased genotype and haplotype inference. Previous studies have shown that haplotypes can be inferred by taking advantage of IGHJ6 heterozygosity, observed in approximately one third of the population. Here, we propose a robust novel method for determining V-D-J haplotypes by adapting a Bayesian framework. Our method extends haplotype inference to IGHD- and IGHV-based analysis, thereby enabling inference of complex genetic events like deletions and copy number variations in the entire population. We generated the largest multi individual data set, to date, of naive B-cell repertoires, and tested our method on it. We present evidence for allele usage bias, as well as a mosaic, tiled pattern of deleted and present IGHD and IGHV nearby genes, across the population. The inferred haplotypes and deletion patterns may have clinical implications for genetic predispositions to diseases. Our findings greatly expand the knowledge that can be extracted from antibody repertoire sequencing data.