GCMM: graph convolution network based on multimodal attention mechanism for drug repurposing

• DOI: 10.1186/s12859-022-04911-8
• Publication Date: 2022-09-13T07:07:13Z
• Authors: Fan Zhang; Wei Hu; Yirong Liu
• Keywords: Drug repositioning; Similarity (geometry)

The main focus of in silico drug repurposing, which is a promising area for using artificial intelligence discovery, the prediction drug-disease relationships. Although many computational models have been proposed recently, it still difficult to reliably predict associations from variety sources data.In order identify potential associations, this paper introduces novel end-to-end model called Graph convolution network based on multimodal attention mechanism (GCMM). In particular, GCMM incorporates known relations, drug-drug chemical similarity, therapeutic disease-disease semantic and target-based similarity into heterogeneous network. A Convolution Network encoder used learn how diseases drugs are embedded various perspectives. Additionally, can enhance performance by applying layer assign levels value features inputting multi-source information.5 fold cross-validation evaluations show that outperforms four recently deep-learning majority criteria. It shows relationships suggests improvement desired metrics. Hyper-parameter analysis exploratory ablation experiments also provided demonstrate necessity each module highest possible level performance. case study Alzheimer's disease (AD). Four five medications indicated correlation coefficient with AD demonstrated through literature or experimental research, demonstrating viability GCMM. All these results imply provide strong effective tool development repositioning.