Degradation Graphs Reveal Hidden Proteolytic Activity

Hartman, Malmström and Wallin (published Feb. 20, 2026) introduce degradation graphs, a probabilistic framework that models proteolysis as directed acyclic networks inferred from single-snapshot peptidomes. They show degradation graphs correct a core bias in conventional quantification, revealing standard methods underestimate upstream proteolytic activity by three- to fourfold across three datasets. The method also yields graph-structured features that improve protease-specific machine learning signatures and interpretability.