Introduction
Accurate drug response prediction (DRP) is essential for advancing precision oncology and drug discovery. However, acquiring labeled DRP data through wet-lab experiments remains costly and time-consuming, resulting in data insufficiency that limits the development of robust computational models. The vast diversity of drug structures and complex tumor heterogeneity exacerbate this challenge, hindering the generalization of models to unseen drugs and new cellular contexts. Therefore, there is an urgent need for a DRP framework with strong generalization ability that can provide reliable predictions for unseen drugs and various biological conditions.
Figure 1. The framework of PVADRP.