Table 1 The AI techniques/tools used in the drug discovery process

Reinforcement learning

Used to optimize drug combinations and dosages by considering multiple interacting variables and maximizing desired outcomes

DeepChem

Open-source library for deep learning in chemistry and drug discovery

DeepTox

Open-source deep learning framework specifically designed for toxicity prediction and assessment

Neural graph fingerprints

Method for encoding molecular structures as fixed-length feature vectors using neural networks, suitable for various applications in drug discovery, such as virtual screening, lead optimization, and property prediction

PotentialNet

Ligand-binding affinity prediction based on a graph convolutional neural network (CNN)

Predictive ADME/Tox modelling

Tools employ ML techniques to model and predict the absorption, distribution, metabolism, excretion, and potential toxicity of drug candidates

Natural language processing (NLP) tools

Assist in extracting and analysing information from scientific literature, patents, and clinical trial data

Cheminformatics tools

Tools enable the analysis and manipulation of chemical structures and properties

QSAR/QSPR modelling

Correlate molecular properties and structures with biological activities or properties, enabling the prediction of compound behaviour

Deep learning (DL)

Applied in tasks like virtual screening, de novo drug design, and predicting drug properties

Machine learning (ML)

Help predict drug-target interactions, analyse biological activity, and optimize lead compounds