基于迁移学习框架的单细胞药物反应预测OA

Tumor tissues contain diverse cell types that differ in morphology,function and biological characteristics.Nota-bly,different cell types have different sensitivity to the same drug,which is also an important reason for the high recurrence rate and low efficacy of cancer treatment.Single-cell RNA sequencing technology is an important research tool for this cellu-lar heterogeneity.This article proposes a transfer learning framework based on variational auto-encoder for single-cell drug response prediction.The framework integrates bulk-sequenced cell line pharmacogenomic data with single-cell RNA se-quencing data,and adopts maximum mean discrepancy and optimal transport to align cell lines and single cells in a low-dimensional latent space,thereby enabling accurate prediction of drug response at the single-cell level.The proposed model was benchmarked against state-of-the-art methods for single-cell drug response prediction and validated on eight data sets,achieving superior predictive performance.Ablation and robustness experiments further verified the effectiveness of each module of the model and the overall stability of the model.This study provides a novel transfer learning approach,offering a new strategy for personalized and precise cancer treatment and contributing to the development of precision medicine.