FastDTI: Drug-Target Interaction Prediction using Multimodality and Transformers

• DOI: 10.7557/18.6788
• Publication Date: 2023-01-24T08:22:54Z
• Authors: Mathijs Boezer; Maryam Tavakol; Zahra Sajadi
• Keywords: Drug target; Benchmark (surveying); Graph Embedding; Multimodality

Recent advances in machine learning have proved effective the application of drug discovery by predicting drugs that are likely to interact with a protein target certain disease, leading prioritizing development and re-purposing efforts. State-of-the-art techniques Drug-Target Interaction (DTI) prediction often computationally expensive can only be trained on small specialized datasets. In this paper, we propose novel architecture, called FastDTI, utilizing pretrained transformers graph neural networks self-supervised manner large-scale (unlabeled) data, which additionally allows for embedding multimodal input representations, both properties. Extensive empirical study demonstrates our approach outperforms state-of-the-art DTI methods KIBA benchmark dataset, while greatly improving computational complexity training, about 200 times faster, excellent performance results.