Estimating the human mutation rate from autozygous segments reveals population differences in human mutational processes

Vagheesh M Narasimhan,Eamonn R Maher,Yali Xue,John Wright,Raheleh Rahbari,Dan Mason,Arthur Wuster,Adam Auton,Matthew E Hurles,Aylwyn Scally,Richard Durbin,Richard C Trembath,Chris Tyler-Smith,David A Van Heel

doi:10.1038/s41467-017-00323-y

Abstract

Heterozygous mutations within homozygous sequences descended from a recent common ancestor offer a way to ascertain de novo mutations across multiple generations. Using exome sequences from 3222 British-Pakistani individuals with high parental relatedness, we estimate a mutation rate of 1.45 ± 0.05 × 10−8 per base pair per generation in autosomal coding sequence, with a corresponding non-crossover gene conversion rate of 8.75 ± 0.05 × 10−6 per base pair per generation. This is at the lower end of exome mutation rates previously estimated in parent–offspring trios, suggesting that post-zygotic mutations contribute little to the human germ-line mutation rate. We find frequent recurrence of mutations at polymorphic CpG sites, and an increase in C to T mutations in a 5ʹ CCG 3ʹ to 5ʹ CTG 3ʹ context in the Pakistani population compared to Europeans, suggesting that mutational processes have evolved rapidly between human populations.

Highlights

Using exome sequences from 3222 British-Pakistani individuals with high parental relatedness, we estimate a mutation rate of 1.45 ± 0.05 × 10−8 per base pair per generation in autosomal coding sequence, with a corresponding non-crossover gene conversion rate of 8.75 ± 0.05 × 10−6 per base pair per generation
Possible shortcomings include: (a) small sample sizes, both in terms of the number of individuals the estimate is obtained from as well as the number of true de novo mutations (DNMs) detected; (b) inaccurate characterization of the false negative (FN) or false positive (FP) rates, perhaps because of comparisons of sequencing data with different properties from different individuals; (c) consideration only of mutations occurring in a single generation, leading to incomplete ascertainment of post-zygotic mutations in parents or offspring[6]; (d) incomplete allowance for the correlation with paternal age; (e) the inclusion of diseased individuals who might have a higher rate of DNMs; or (f) failure to account for gene conversion events
We restricted our analysis to autosomal single-nucleotide substitutions with the same genotype call from both samtools[10] and Genome Analysis Toolkit (GATK)[11] when calling across all samples

Summary

Introduction

Using exome sequences from 3222 British-Pakistani individuals with high parental relatedness, we estimate a mutation rate of 1.45 ± 0.05 × 10−8 per base pair per generation in autosomal coding sequence, with a corresponding non-crossover gene conversion rate of 8.75 ± 0.05 × 10−6 per base pair per generation. Possible shortcomings include: (a) small sample sizes, both in terms of the number of individuals the estimate is obtained from as well as the number of true de novo mutations (DNMs) detected; (b) inaccurate characterization of the false negative (FN) or false positive (FP) rates, perhaps because of comparisons of sequencing data with different properties from different individuals; (c) consideration only of mutations occurring in a single generation, leading to incomplete ascertainment of post-zygotic mutations in parents or offspring[6]; (d) incomplete allowance for the correlation with paternal age; (e) the inclusion of diseased individuals who might have a higher rate of DNMs; or (f) failure to account for gene conversion events To address these shortcomings, and to obtain an estimate which, like population-genetic approaches, averages over multiple generations and many mutational events, we adopted an approach based on observing heterozygous genotypes within sequence intervals inherited identical-by-descent (IBD) from a recent common ancestor (autozygous segments). As our estimate is one of the few data sets of DNMs obtained in a non-European population we examine differences in context-specific mutational spectra between human populations

Methods

Results

Conclusion