Simulation of ENU mutagenesis

Simulation of ENU mutagenesis was accomplished by using breeding and genotype data from the G0, G1, and G2 mice available in the Mutagenetix database (retrieved March 2017; http://mutagenetix.utsouthwestern.edu); G3 mice were “bred” in silico by a random process. To determine the number of G3 mice in each pedigree when all genes were hypothetically neutral, we fit a linear regression model of pedigree sizes regressed by pedigree type, number of litters produced, and number of each type of mutations in each G2 mother (Supplementary Fig. ⁵). Then, all mutation counts were set to zero, which is equal to the assumption that all mutations in the genes were neutral. Next, the simulation randomly generated G3 mice and their genotypes from in silico “mating” based on the genotypes of the G2 mother and father, and the hypothetical pedigree sizes. Then, an “essential” quality was randomly assigned to a varying fraction of genes, and all mutations in such genes were designated as truly damaging or not based on their PP2 mutation category and the corresponding estimated damage probabilities. The assigned essential genes were further sampled to be “partially essential” or “totally essential” with a ratio of 39:100. For each simulated percentage of essential genes, G3 mice were computationally “culled” if they were homozygous for designated truly damaging mutations in designated totally essential genes; G3 mice homozygous for designated truly damaging mutations in designated partially essential genes were culled at a rate of 50%.