vamos: VNTR annotation using efficient motif sets

MotivationRoughly 3% of the human genome is composed of variable-number tandem repeats (VNTRs): tandemly repeated arrays of motifs at least six bases. These loci are highly polymorphic: over 61% of insertion and deletion variants at least 50 bases found from long-read assemblies are inside VNTRs. Furthermore, long-read assemblies reveal that VNTR loci are multiallelic, and can vary by both motif composition and copy number. Current approaches that define and merge variants based on alignment breakpoints do not capture this complexity of variation. A natural alternative approach is to instead define the motif composition of VNTR sequences from samples, and to detect differences based on comparisons of repeat composition. However, due to the complexity of VNTR sequences, it is difficult to establish a common reference set of motif sequences that may be used to describe variation in large sequencing studies. ResultsHere we present a method vamos: VNTR Annotation using efficient Motif Sets that for any VNTR locus selects a set of representative motifs from all motifs observed at that locus that may be used to encode VNTR sequences within a bounded edit distance of the original sequence. We use our method to characterize VNTR variation in 32 haplotype-resolved human genomes. In contrast to current studies that merge multi-allelic calls, we estimate an average of 3.1-4.0 alleles per locus. Availabilitygithub.com/chaissonlab/vamos, zenodo.org/record/7158427 Contactmchaisso@usc.edu