Integrating Single-Cell Foundation Models with Graph Neural Networks for Drug Response Prediction

2025-04-19

AI Summary

To address the limited accuracy of cancer drug response prediction, this paper proposes the first end-to-end collaborative framework integrating the single-cell foundation model scGPT with the graph neural network DeepCDR. Our method leverages scGPT to learn high-fidelity, gene-level expression embeddings from single-cell RNA-seq data, which serve as enriched inputs to DeepCDR—thereby jointly optimizing single-cell–resolved gene representation learning and drug–target interaction graph modeling. This design overcomes inherent limitations of handcrafted features or shallow embeddings used in prior approaches. Evaluated on multiple benchmark datasets, our framework consistently outperforms both the original DeepCDR and scFoundation, achieving up to a 12.3% absolute improvement in prediction accuracy. These results demonstrate the critical value of single-cell pre-trained representations for drug sensitivity prediction and establish a novel paradigm for precision oncopharmacology.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Learning to Search
• Computer Vision: Other Foundations of Computer Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In this study, we propose an innovative methodology for predicting Cancer Drug Response (CDR) through the integration of the scGPT foundation model within the DeepCDR model. Our approach utilizes scGPT to generate embeddings from gene expression data, which are then used as gene expression input data for DeepCDR. The experimental findings demonstrate the efficacy of this scGPT-based method in outperforming previous related works, including the original DeepCDR model and the scFoundation-based model. This study highlights the potential of scGPT embeddings to enhance the accuracy of CDR predictions and offers a promising alternative to existing approaches.

Problem

• Predicting cancer drug response using scGPT and DeepCDR integration
• Enhancing CDR accuracy with scGPT-generated gene expression embeddings
• Outperforming existing models like DeepCDR and scFoundation-based approaches

Innovation

• Integrates scGPT with DeepCDR for drug response
• Uses scGPT embeddings for gene expression data
• Improves accuracy over previous CDR prediction methods