High rate of mutation and efficient removal by selection of structural variants from natural populations of .

The importance of genomic structural variants (SVs) is well-appreciated, but much less is known about their mutational properties than of single nucleotide variants (SNVs) and short indels. The reason is simple: the longer the mutation, the less likely it will be covered by a single sequencing read, thus the harder it is to map unambiguously to a unique genomic location. Here we report SV mutation rate estimates from six mutation accumulation (MA) lines from two strains of (N2 and PB306) using long-read (PacBio) sequencing. The inferred SV mutation rate 1/10 the SNV rate and ~1/4 the short indel rate. We identified 40 mutations, and removed 52 false positives (FP) by manual inspection of each SV call. Excluding one atypical line (5 mutations, 35 FPs), the signal (inferred mutant) to noise (FP) ratio is approximately 2:1. False negative rates were determined by simulating variants in the reference genome, and observing 'recall'. Recall rate ranges from >90% for short indels and declines as SV length increases. Small deletions have nearly the same recall rate as small insertions (100bp), but deletions have higher recall rates than insertions as size increases. The reported SV mutation rate is likely an underestimate. A quarter of identified SV mutations occur in SV hotspots that harbor pre-existing low complexity repeat variation. By comparison of the spectrum of spontaneous SVs to wild isolates, we infer that natural selection is not only efficient at removing SVs in exons, but also removes roughly half of SVs in intergenic regions.