Machine learning in chemoinformatics and drug discovery

Abstract

Chemoinformatics is an established discipline focusing on extracting, processing and extrapolating meaningful data from chemical structures. With the rapid explosion of chemical 'big' data from HTS and combinatorial synthesis, machine learning has become an indispensable tool for drug designers to mine chemical information from large compound databases to design drugs with important biological properties. To process the chemical data, we first reviewed multiple processing layers in the chemoinformatics pipeline followed by the introduction of commonly used machine learning models in drug discovery and QSAR analysis. Here, we present basic principles and recent case studies to demonstrate the utility of machine learning techniques in chemoinformatics analyses; and we discuss limitations and future directions to guide further development in this evolving field.

Figure 1

Computational workflow for chemoinformatics analysis using machine learning. The first step of chemoinformatics analysis is feature extraction, through which the compound is characterized by substructure fragments or other chemical descriptors (first box). The chemical features of the compound are represented by chemical fingerprints and applied for compound similarity comparison based on the presence and absence of shared chemical features. The chemical fingerprint can be used for predicting other chemical and physiochemical properties in QSAR/QSPR analysis using diverse machine learning models including making inference from the training data by comparison (instance-based learning) or from the trained statistical model (model-based learning) (second box).