Additional file 2 of MethylationToActivity: a deep-learning framework that reveals promoter activity landscapes from DNA methylomes in individual tumors

Additional file 2: Table S1. H3K27ac active cancer consensus genes in 3 NBL cell lines, and 3 NBL O–PDX samples. Table S2. Baseline models vs. vanilla M2A predictive performance comparison. Table S3. M2A predictive performance in NBL cell line samples. Table S4. Observed H3K27ac and H3K4me3 ENCODE replicate consistencies. Table S5. M2A predictive performance in RMS O–PDX samples. Table S6. M2A RMS transfer model predictive performance in RMS O–PDX samples. Table S7. M2A predictive performance in ENCODE dataset. Table S8. M2A predictive performance in AML samples. Table S9. ERMS vs. ARMS DMRs and associated genes (overexpressed in ERMS). Table S10. ERMS vs. ARMS DMRs and associated genes (overexpressed in ARMS). Table S11. M2A alternate promoter usage predictive performance between ARMS and ERMS samples. Table S12. M2A predictive performance in EWS samples, before and after transfer. Table S13. A univariate survival analysis of differential H3K27ac promoter activity between TP53 mutant and TP53 wild–type EWS tumors. Table S14. A univariate survival analysis of alternate promoter usage between TP53 mutant and TP53 wild–type EWS tumors. Table S15. ENCODE H3K27ac replicate consistency with gene expression. Table S16. Dataset availability. Table S17. M2A model topologies. Table S18. Parameter tuning: mean performance in the NBL validation set (R2). Table S19. Sample summary information.