AI model predicts metastasis (MangroveGS)
MangroveGS, an AI that analyzes gene‑expression patterns, predicts colon tumor metastatic potential with ~80% accuracy—pointing to a future where molecular data and cytology feed prognostic models. That raises questions about integrating predictive analytics into multidisciplinary tumor boards. (sciencedaily.com)
The paper "Emergence of high‑metastatic potentials and prediction of recurrence and metastasis" appears in Cell Reports (Vol. 45, Issue 1) under DOI 10.1016/j.celrep.2025.116834 and lists Aravind Srinivasan and Ariel Ruiz i Altaba among the lead authors. (sciencedirect.com) Researchers derived transcriptomes from roughly 30 single‑cell‑derived clones taken from two primary colon tumors, then correlated in vitro migration assays and mouse lung‑metastasis readouts with RNA‑seq to identify genes that track metastatic potential. (news-medical.net) The team defined Metastatic Potential Gradient Genes (MPGGs) and built MangroveGS using gene‑signature ensembles and ensemble strategies called CSE and MPGGs‑TOP, then validated the model across TCGA and independent clinical cohorts where it outperformed existing molecular signatures and staging systems. (cell.com) Functional perturbation experiments reported that CRISPR knockout of CSAG1 or ATP11C reduced high‑MP cell migration and in vivo metastasis while VGLL1 perturbation altered migration in the opposite direction, and implicated ER stress, polycomb/SWI‑SNF epigenetic regulators, and lipid mediators (LPA/Autotaxin) as mechanistic contributors. (sciencedirect.com) The authors state MangroveGS was trained on cloned‑cell transcriptomes but applied to bulk tumor datasets for pan‑cancer prediction, with reported applicability to pancreas, bladder, stomach, cervical, lung, prostate and breast epithelial cancers. (technologynetworks.com) The manuscript highlights candidate anti‑metastatic intervention points including Polycomb components, CSAG1 and ATP11C and frames MangroveGS as a platform both for prioritizing therapeutic targets and for stratifying patients in future recurrence/metastasis‑directed trials. (sciencedirect.com)
