Long-read sequencing of trios reveals increased germline and postzygotic mutation rates in repetitive DNA

Long-read sequencing (LRS) has improved sensitivity to discover variation in complex repetitive regions, assign parent-of-origin, and distinguish de novo germline from postzygotic mutations (PZMs). Most studies have been limited to population genetic surveys or a few families. We applied three orthogonal sequencing technologies-lIlumina, Oxford Nanopore Technologies, and Pacific Biosciences-to discover and validate de novo mutations (DNMs) in 73 children from 42 autism families (157 individuals). Assaying 2.77 Gbp of the human genome using read-based approaches, we discover on average 95 DNMs per transmission (87.5 de novo single-nucleotide variants and 7.8 indels), including sex chromosomes. We estimate that LRS increases DNM discovery by 20-40% over previous Illumina-based studies of the same families, and more than doubles the discoverable number of PZMs that emerged early in embryonic development. The strict germline mutation rate is 1.30×10^-8 substitutions per base pair per generation, strongly driven by the father's germline (3.95:1), while PZMs increase the rate by 0.23×10^-8 with a modest but significant bias toward paternal haplotypes (1.15:1). We show that the mutation rate is significantly increased for classes of repetitive DNA, where segmental duplication (SD) mutation shows a dependence on the length and percent identity of the SD. We find that the mutation rate enrichment in repeats occurs predominantly postzygotically as opposed to in the germline, a likely result of faulty DNA repair and interlocus gene conversion.