EXPRESSO predicts cancer treatment response from tumor transcriptomics

Researchers developed EXPRESSO (EXpression Profile RESponSe Optimizer), a supervised machine-learning framework that integrates drug targets and context-specific biomarkers to predict treatment response from pretreatment tumor transcriptomes. The resource included 91 cohorts comprising 5,675 patients with nine cancer types treated with six first-line therapies: immune checkpoint inhibitors targeting PD-1/PD-L1, trastuzumab, bevacizumab, BRAF inhibitors, paclitaxel, and FAC/FEC chemotherapy.

The authors report that EXPRESSO outperformed 20 published transcriptomic signatures and other machine-learning methods. Validation across 22 independent cohorts supported generalizability beyond cross-validation, and the signature also stratified progression-free survival in cohorts treated with immune checkpoint blockade. Predictive performance plateaued with increasing training data for some therapies but continued to improve for others. The findings are promising for transcriptomic response modeling, but the abstract does not provide performance estimates or evidence of prospective clinical utility; the framework is not established as a basis for treatment decisions.