Novel Consensus Architecture To Improve Performance of Large-Scale Multitask Deep Learning QSAR Models

• DOI: 10.1021/acs.jcim.9b00526
• Publication Date: 2019-10-04T15:45:42Z
• Authors: Alexey V. Zakharov; Tongan Zhao; Dac-Trung Nguyen; Tyler Peryea; Timothy Sheils; Adam Yasgar; Ruili Huang; Noel Southall; Anton Simeonov
• Keywords: chEMBL; Chemical space; PubChem; Transfer of learning

Advances in the development of high-throughput screening and automated chemistry have rapidly accelerated production chemical biological data, much them freely accessible through literature aggregator services such as ChEMBL PubChem. Here, we explore how to use this comprehensive mapping biology space support large-scale quantitative structure–activity relationship (QSAR) models. We propose a new deep learning consensus architecture (DLCA) that combines multitask approaches together generate QSAR This method improves knowledge transfer across different target/assays while also integrating contributions from models based on descriptors. The proposed approach was validated compared with proteochemometrics, learning, Random Forest methods paired various descriptors types. DLCA demonstrated improved prediction accuracy for both regression classification tasks. best their modeling sets are provided publicly available web at https://predictor.ncats.io.