CSGDA: A Cell State-Guided Graph Domain Adaptation Network for Single-Cell Drug Response Prediction.

CSGDA achieves approximately 6 percentage points higher accuracy and AUPR than the previous best method in predicting single-cell drug responses across diverse biological conditions.

• Why it matters: Understanding and predicting drug responses at the single-cell level is crucial for tackling tumor heterogeneity, metastasis, and resistance, which are major challenges in precision cancer therapy.
• What they did: The approach integrates biological priors to map gene expression into cell states, constructs cell topology via structure learning, and employs graph domain adaptation with an overlap penalty, tested on five scRNA-seq datasets.
• The result: CSGDA not only improves prediction performance but also identifies key resistance genes, advancing the potential for personalized treatments and better understanding of tumor heterogeneity.