AI and the World's Hard Problems · Medicine & drug discovery · Nature (computational biology) · 2026-10-26
Interpretable machine learning models for oncotypedDX® prediction using clinicopathologic data: a brazilian cohort study
Oncotype DX guides chemotherapy decisions in early breast cancer, but its high cost limits access in resource-constrained settings. Using 935 patients from a Brazilian cohort, the authors built an interpretable ridge regression model from routinely available clinicopathologic variables to predict whether the recurrence score exceeds 25. The model reached an AUROC of 0.818, and its highest-risk quartile contained 60.8% of cases with a score above 25; the authors suggest up to 75% of patients could be triaged without confirmatory testing.