Related Experiment Video
Updated: May 19, 2026
Implementation of In Vitro Drug Resistance Assays: Maximizing the Potential for Uncovering Clinically Relevant Resistance Mechanisms
Published on: December 9, 2015
Weak supervision of H&E slides reveals systems-level biology and functional states that govern therapeutic resistance
Tiago Goncalves1,2, Dagoberto Pulido2, Carmen M Perrino2
1INESC TEC, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal.
Abstract:
Precision oncology lacks scalable tools to assess, at the patient level, systems-level tumor microenvironment (TME) programs driving therapeutic resistance. To address this gap, we trained a weakly-supervised deep learning model, using routine H&E slides as input, to derive quantitative activity for therapeutically-relevant TME phenotypes, spanning immune, metabolic, and tumor cell-intrinsic programs. Using 3111 breast cancer H&E WSIs with matched bulk transcriptomics, our model accurately infers these biological states (AUROC>0.80; PCC>0.64). Validation spanned three levels: (i) tissue-matched multiplexed immunofluorescence, showing concordance between inferred functional states and immune cell fractions (p=0.006-0.106), (ii) blinded reader assessments, confirming localization of phenotype-specific morphology (p<3×10-5), and (iii) multi-institutional patient cohorts, where model-derived phenotypes stratified for clinical response (p<0.045). Despite relying on slide-level labels for training, our model's attention mechanism identifies focal regions of tumor tissue that drive the overarching clinical phenotype or treatment response. By extracting spatially resolved TME biology from routine histology, this strategy can be applied to massive legacy biobanks to enable discovery of new morpho-molecular mediators of resistance across real-world patient populations.
Insights
This study introduces a deep learning model using H&E whole-slide images to analyze tumor microenvironment phenotypes and predict therapeutic resistance in cancer patients. The tool offers a scalable method to assess complex biological programs from routine pathology slides.
Area of Science:
- Computational Biology
- Pathology
- Oncology
Background:
- Precision oncology requires tools to analyze the tumor microenvironment (TME) and understand therapeutic resistance.
- Current methods for TME analysis are often not scalable or require resource-intensive data.
Purpose of the Study:
- To develop a scalable deep learning model to assess TME phenotypes from routine H&E whole-slide images (WSIs).
- To quantitatively derive activity of immune, metabolic, and tumor cell-intrinsic programs driving therapeutic resistance.
Main Methods:
- Trained a weakly-supervised deep learning model on 3111 breast cancer H&E WSIs with matched bulk transcriptomics.
- Inferred pathway enrichment scores representing TME phenotypes.
- Validated model performance using multiplexed immunofluorescence, blinded reader assessments, and multi-institutional patient cohorts.
Main Results:
- The model accurately inferred TME biological states (AUROC>0.80; PCC>0.64).
- Inferred functional states showed concordance with immune cell fractions (p=0.006-0.106).
- Model-derived phenotypes stratified clinical response in patient cohorts (p<0.045).
Conclusions:
- This deep learning approach provides a scalable method to assess functional TME biology from widely available H&E WSIs.
- It complements spatial Omics by leveraging routine pathology data and clinical outcomes for therapeutic resistance research.
- The strategy enables pan-cancer investigation of TME programs using legacy biobanks.
More Related Videos
08:59Looking for Driver Pathways of Acquired Resistance to Targeted Therapy: Drug Resistant Subclone Generation and Sensitivity Restoring by Gene Knock-down
Published on: December 11, 2017
08:52Profiling Sensitivity to Targeted Therapies in EGFR-Mutant NSCLC Patient-Derived Organoids
Published on: November 22, 2021
