Transcriptome-informed multi-modal AI for predicting neoadjuvant therapy response from breast cancer biopsies
Two-stage model infers transcriptome from breast cancer histopathology to predict neoadjuvant therapy response, reaching pooled AUROC 0.79 across nine cohorts.
The first stage learns transcriptome prediction from histopathology using 8,742 patients across 32 cancer types; the second stage predicts pathological complete response from inferred expression plus clinical variables. Trained on 1,080 patients (five cohorts) and evaluated on 1,412 patients (nine cohorts), the model achieves pooled AUROC 0.79 (95% CI 0.73-0.85) and discriminates responders within molecular subtypes. Ablations show transcriptome-wide inference outperforms clinical variables alone and one-stage pathology models.
- Two-stage design: infer transcriptome from pathology, then predict therapy response
- Pooled AUROC 0.79 across nine external evaluation cohorts
- Outperforms histopathological biomarkers with minimal biopsy tissue
- Robust across intratumoral sampling variation
From the original article
1 min read at arxiv.orgScarcity of labeled data limits development of deep learning biomarkers in oncology. We 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.
By Jungkyu Park · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.