Deep multiple instance learning on heterogeneous graph for drug-disease association prediction

Yaowen Gu, Si Zheng, Bowen Zhang, Hongyu Kang, Rui Jiang, Jiao Li

Abstract: Highlights•Deep Multiple Instance Learning is introduced in drug repositioning for the first time.•Meta-path instances is elaborately designed and efficiently represented in proposed model MilGNet.•MilGNet showed noteable improvements in 5 benchmarks compared with 8 baselines (10.9% on AUPR).•Three case studies emphasize potentials of MilGNet in interpretable indication and therapy discovery.