Related Experiment Video
Updated: May 15, 2026
03:37
Generating the Transcriptional Regulation View of Transcriptomic Features for Prediction Task and Dark Biomarker Detection on Small Datasets
Published on: March 1, 2024
Multimodal interpretable deep learning for transcriptome-informed precision oncology and drug mechanism analysis.
Ning Qu1,2, Xiaochu Tong1,2, Zhaokun Wang1,2
1Drug Discovery and Design Center, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, 555 Zuchongzhi Road, Shanghai, China.
NPJ Digital Medicine
|May 13, 2026
Summary
BioGDR, a new deep learning framework, enhances precision oncology by predicting drug responses using integrated biological data. It offers mechanistic insights and outperforms existing methods in drug discovery and clinical applications.
Area of Science:
- Computational biology
- Bioinformatics
- Machine learning in oncology
Background:
- Precision oncology requires interpreting complex cellular signals for predicting drug responses in diverse cancer types.
- Current methods often lack the ability to integrate multimodal data for comprehensive analysis.
Purpose of the Study:
- To introduce BioGDR, a multimodal interpretable deep learning framework for precision oncology.
- To enable mechanistic insights into drug sensitivity by integrating biological features without experimental measurements.
Main Methods:
- Utilized pathway-informed graph neural networks to model tumor transcriptomic states.
- Employed a drug-guided attention strategy for multimodal data integration.
- Integrated structure-based predicted biological features like differential gene expression and kinase inhibition profiles.
Main Results:
- BioGDR demonstrated superior performance in compound screening for early-stage drug discovery.
- Achieved high accuracy in predicting cell line sensitivity across heterogeneous cellular states.
- Clinical patient cohort analyses confirmed BioGDR's practical utility and generalization capability.
- Experimental validation identified sensitive cell populations and elucidated mechanisms for a novel ALDH1B1 inhibitor.
Conclusions:
- BioGDR provides a robust, biologically informed framework for advancing precision oncology.
- The framework effectively bridges preclinical drug development and clinical applications.
- Integrative, multimodal learning and interpretable mechanism analysis are key to BioGDR's success.
