Signature scoring in bulk RNA-seq from Abida et al. cohort

Clinical annotations and expression quantifications were obtained from the published supplementary materials and from the authors directly⁸. We converted gene expression values from fragments per kilobase of transcript per million mapped reads (FPKM) (as supplied by the authors) to TPM for consistency with the rest of our study. As this cohort included samples sequenced at different centers and from different metastatic sites, we further restricted our analyses to avoid batch effects. For Fig. 2d, we analyzed only samples sequenced via transcriptome capture at the University of Michigan, as this was the largest identifiably uniformly sequenced subset. We could not analyze bone biopsies in this cohort due to paucity of post-enzalutamide samples. For Fig. 3c, as the largest number of small cell carcinoma samples were sequenced at Cornell, we included only small cell carcinoma and adenocarcinoma cases from Cornell in our analyses.