Fig. 7

Genomic and epigenomic features influencing SNV and SV rates at loop anchor points in breast tumours. a, b Variable importance, measured as percentage increase in mean squared error (%IncMSE) when each predictor variable is removed from the model, predicting SNV rates (a) or SV breakpoint rates (b), respectively. c Spearman’s rank correlation coefficient estimates between mutation rates and predictor variables of the random forest models, for SNV rates (green) and SV breakpoint rates (blue); 95% confidence intervals of the estimates are indicated. d SNV rates in BRCA-EU breast tumours within 500 kb of LAPs in the MCF-7 breast tumour cell line relative to replication timing (average MCF-7 ENCODE Repli-seq signal); higher Repli-seq values indicate earlier replication time