Motivation Cancer progression is driven by the stochastic accumulation of interacting genetic alterations, and tumors from different tissues can share common evolutionary patterns despite distinct anatomical origins. While simplified progression models can be applied at scale, the use of expressive models that capture complex inter-event dependencies remains challenging in pan-cancer settings due to their computational demands.

Results We present fastMHN, a scalable approximation to Mutual Hazard Network-based cancer progression models, that enables pan-cancer analysis of large genomic datasets with the explicit aim of enabling transfer of progression mechanisms across tumor entities. Applying fastMHN to a large clinical cohort, we identify a tumor-progression group spanning multiple tissues that is characterized by a link between STK11 mutations and poor patient survival. While the clinical relevance of STK11 mutations is well established in non-small cell lung cancer, our results suggest that a similar progression mechanism is present in molecularly defined subgroups of other cancer types. These findings illustrate how scalable pan-cancer progression modeling can facilitate cross-entity transfer of biological and potentially clinical insights.