3D graph neural network with few-shot learning for predicting drug-drug interactions in scaffold-based cold start scenario

Qiujie Lv, Jun Zhou, Ziduo Yang, Haohuai He, Calvin Yu-Chian Chen

Abstract: Highlights•We develop a competitive tool for DDI prediction in cold start scenario.•We design a 3DGNN that continuously model on discretized atoms and incorporates invariant descriptors.•We utilize a few-shot learning strategy to transfer meta-knowledge from existing drugs to new drugs.•We design scaffold-based cold start scenarios to evaluate the generalization performance of DDI model.