a) Distribution of P-values (log10 transformed) from a randomized test of PGLS models for association between aggregate gene count and malignancy rate. Simulated gene sets replicated the gene set “negative regulation of transforming growth factor beta production” by randomly sampling 12 genes with similar variance to those in the set. The arrow shows the actual P-value of the PGLS model for this set. b) Distribution of P-values (log10 transformed) from 1,000 iterations of a simulation test of PGLS between gene count of set “negative regulation of transforming growth factor beta production” and a random continuous phenotype simulated under Brownian motion. The arrow shows the actual P-value of the PGLS model with malignancy rate as the phenotype.