Hereditary cancer genes are highly susceptible to splicing mutations
Fig 5
Random forest classification and prediction of SSM-prone genes.
A. The order of variable importance by mean decease in accuracy for SSM-prone genes versus genes with an expected number of SSM. The directions that associate with SSM-prone genes are indicated, positive directions are green, and negative directions are red. B. Classification performance of the random forest models and the logistic regression models was calculated as the area under the curve (AUC) in receiver operating characteristic (ROC) analysis. C. Scheme of random forest classification on all genomic genes. D. Average proportion of low frequency ExAC splice-site variants per splice-site in predicted SSM-prone genes (probability: 0.60–0.86) versus genes not predicted to be SSM-prone (P = 6.1043e-18, Mann-Whitney). E. Common variants are depleted from the category of variants that cause loss of splice-site signal at the 5′ splice-site (upper plot). Rare variants are enriched in the range of the splice site signal scores that abolish 5′ splice-site recognition (lower plot).