Additionally, we utilized the Bayesian information criterion (BIC) for model selection, implemented in R using the bic.glm function from the BMA package. Based on the BIC, the best model included percent of bases ≥ 2× and maternal age only. However, we chose to utilize all of the covariates identified with the AIC because we felt it was important to include additional covariates that may be more relevant when considering other classes of de novo mutation. Nonetheless, Poisson regression with the simplified model above results in consistent results (not shown).
