Paper Overview
Field: Machine Learning Posted: 2026-10-02 arXiv: 2610.03693
Authors: Jungkyu Park, Dhruva Biswas, Joseph Cappadona, Cerise Tang, Ken G. Zeng, Bartosz Machura, Chuwen Liu, Paolo Tarantino, Coral Omene, Francisco J. Esteva, Rohit Bhargava, Marcin Braun, Kamila Paździerz, Jakub Czerwiński, Hanna Romańska-Knight, Albert Grinshpun, Bareket Daniel, Michele Buchinger, Frederick Howard, Piotr Wysocki, Brie Chun, Freya Schnabel, Rich Caruana, Jan Witowski, Krzysztof J. Geras
Abstract
Scarcity of labeled data limits development of deep learning biomarkers in oncology. The authors develop a two-stage AI model predicting pathological complete response (pCR) to neoadjuvant therapy in breast cancer. The first stage learns the transcriptome from histopathology using 8,742 patients across 32 cancer types, corroborated by pathologist review and spatial agreement with measured expression. This simplifies the second stage to predicting pCR from inferred expression and clinical variables.
Developed using 1,080 patients (five cohorts) and evaluated in 1,412 patients (nine cohorts), the model achieves a pooled AUROC of 0.79 (95% CI, 0.73–0.85), discriminating responders within molecular subtypes. It outperforms histopathological biomarkers, remaining stable across intratumoral sampling and limited biopsy tissue. Ablation studies show that whole-transcriptome inference provides better discrimination than clinical variables alone or a single-stage pathology model, and enhances robustness by avoiding the gene-selection constraints of genomic assays.
Key Findings
- Two-stage design: histopathology → inferred transcriptome → pCR prediction with clinical variables
- Pooled AUROC 0.79 (95% CI, 0.73–0.85) across nine external evaluation cohorts
- Works within molecular subtypes, distinguishing responders
- More robust than single-stage models and genomic assays due to whole-transcriptome inference
- The authors suggest this biology-informed compression approach may generalize to other data-sparse precision oncology tasks