Cancer drug response prediction with surrogate modeling-based graph neural architecture search

• DOI: 10.1093/bioinformatics/btad478
• Publication Date: 2023-08-08T12:44:53Z
• Authors: Babatounde Moctar...; Jianliang Gao; Jiamin Chen; Raeed Al-Sabri; Zhenpeng Wu
• Keywords: Benchmark (surveying); Hyperparameter

Understanding drug-response differences in cancer treatments is one of the most challenging aspects personalized medicine. Recently, graph neural networks (GNNs) have become state-of-the-art methods many representation learning scenarios bioinformatics. However, building an optimal handcrafted GNN model for a particular drug sensitivity dataset requires manual design and fine-tuning hyperparameters model, which time-consuming expert knowledge.In this work, we propose AutoCDRP, novel framework automated predictor using GNNs. Our approach leverages surrogate modeling to efficiently search effective architecture. AutoCDRP uses predict performance architectures sampled from space, allowing it select architecture based on evaluation performance. Hence, can identify by exploring all space. Through comprehensive experiments two benchmark datasets, demonstrate that generated surpasses designs. Notably, identified consistently outperforms best baseline first epoch, providing further evidence its effectiveness.https://github.com/BeObm/AutoCDRP.