Cancer evolves through the emergence and selection of molecular alterations. Cancer genome profiling has revealed that specific events are more or less likely to be co-selected, suggesting that the selection of one event depends on the others. However, the nature of these evolutionary dependencies and their impact remain unclear. Here, we designed SELECT, an algorithmic approach to systematically identify evolutionary dependencies from alteration patterns. By analyzing 6,456 genomes from multiple tumor types, we constructed a map of oncogenic dependencies associated with cellular pathways, transcriptional readouts, and therapeutic response. Finally, modeling of cancer evolution shows that alteration dependencies emerge only under conditional selection. These results provide a framework for the design of strategies to predict cancer progression and therapeutic response.

Hi-C chromatin maps of EZH2 WT and EZH2 Y646X lymphomas. EZH2 p.Tyr646* gain-of-function alterations lead to a genomewide increase in H3K27me3 (Supplementary Fig. 1a). To establish whether this global accumulation of H3K27me3 modifies the genome topology on a similarly broad scale, we performed high-throughput chromatin conformation capture (Hi-C) in two lymphoma cell lines (Karpas422 and WSU-DLCL2) expressing the mutant form of EZH2 (EZH2 Y646X) and a lymphoma cell line (OCI-Ly19) expressing the wild-type EZH2 (EZH2 WT) protein (Supplementary Table 1 and Supplementary Note). Contact maps of EZH2 Y646X and EZH2 WT cells were binned in regions of 50 kb and compared with multiple metrics (Fig. 1). For each pair of maps, we compared the overall distribution of chromosomal contacts by using the stratum-adjusted correlation coefficient (SCC) 24 (Fig. 1a); the fraction of 1-Mb regions assigned to the same compartment (A or B) 3 (Fig. 1b); the similarity among TADs 25-27 (Fig. 1c); and the fraction of bin interactions that were significant in both maps, also known as the cell interactome 28,29 (Fig. 1d). To build a reference scale of values for each metric, we compared Hi-C maps of EZH2 Y646X lymphoma cell lines with Hi-C maps of endothelial cells (HUVEC), fetal fibroblasts (IMR90), and normal lymphoblastoid cells (GM12878). Moreover, we used randomized contact maps or