The Usage of Artificial Intelligence in Drug Discovery

Written by William Hsieh, PharmD 2025

January 6, 2024

The Usage of Artificial Intelligence in Drug Discovery

Drug discovery in the modern era has relied on large computational arrays sifting through numerous chemical compounds, seeking to see a match to a particular enzymatic receptor and to analyze drug interactions. Structure-activity relationship (SAR) studies frequently incorporate biophysical interactions that are rigorously defined within the framework of quantum physics. Furthermore, computational chemistry plays a significant role in influencing these parameters. An example of an algorithm commonly used in these calculations is Monte Carlo simulations, which are often used to compute proteins' pathways and thermodynamic properties. As a result of the sheer volume of data and pattern recognition involved in this process, artificial intelligence has stepped up to help revolutionize drug discovery.

Wong et al. detail how his team discovered a new structural class of antibiotics using AI. This is particularly important in light of the growing issue of antibiotic resistance, of which Methicillin-resistant Staphylococcus aureus (MRSA) is an example. The study uses a substructure-based approach, chemical analysis, and graph neural networks to estimate how potential antibiotics would work and how harmful they would be to cells in millions of compounds. The method was tested on various bacteria, including MRSA and vancomycin-resistant enterococci, leading to the discovery of a new structural class of antibiotics effective against these strains. Notably, this novel drug class was selective for MRSA, showed low resistance potential, and proved effective in mouse models. The study developed models to predict the cytotoxicity of compounds in human cells using a dataset of 39,312 compounds. These models, varying in predictive accuracy across cell types, aided in identifying compounds effective against S. aureus with minimal human cell toxicity. Once the model was trained to acceptable levels, graph neural network ensembles predicted antibiotic activity and cytotoxicity for over 12 million compounds. These algorithms were crucial in identifying promising compounds and demonstrating a new route for combating antibiotic resistance.1

These practices have been utilized during the battle against COVID-19 during the pandemic. The repurposing of conventional medications (such as the drug combination of hydroxychloroquine, azithromycin, and tocilizumab) has limitations for treating COVID-19, as Dr. Ho discusses in the study "Addressing COVID-19 Drug Development with Artificial Intelligence." Thus, the focus is shifted toward utilizing AI to explore the vast parameter space of potential drug combinations and dosages. Specifically, Project IDentif.AI has demonstrated potential for enhancing combination therapy using a neural network-based methodology. This approach relies on establishing a quadratic correlation between inputs, which consist of drugs and their corresponding doses, and outputs, which are defined by treatment efficacy and safety. Consequently, this development has the potential to improve the effectiveness and safety of treatments, thereby serving as a valuable asset in the preparation for future pandemics by enabling the implementation of more precise and efficient therapeutic approaches.2

The study by Jayatunga et al. emphasized the substantial influence of artificial intelligence (AI) on small-molecule drug discovery. Pharmaceutical companies that utilize AI have shown rapid pipeline growth, notably in well-known target areas like oncology. A specific example is TYK2 inhibitors, where AI identified a novel, highly selective compound, demonstrating AI's potential to discover unique molecules. The overall effect of AI, however, varies. While there are early signs of improved discovery efficiency and the possibility of novel chemical compounds, it is still unclear how long-term clinical success and cost implications will pan out.3

The eventual incorporation of artificial intelligence has the potential to revolutionize drug discovery, marking a paradigm shift in tackling diseases. From predicting antibiotic activity to optimizing drug combinations, AI's role demonstrates its invaluable contribution to pharmaceutical research. While the long-term outcomes and cost implications remain fully understood, AI's current impact, particularly in small-molecule drug discovery and repurposing of existing drugs, signifies a promising future in efficient and effective healthcare solutions.

References:

1. Wong F, Zheng EJ, Valeri JA, et al. Discovery of a structural class of antibiotics with explainable deep learning. Nature. Published online December 20, 2023:1-9. doi:https://doi.org/10.1038/s41586-023-06887-8

2. Ho D. Addressing COVID-19 Drug Development with Artificial Intelligence. Advanced Intelligent Systems. Published online April 10, 2020:2000070. doi:https://doi.org/10.1002/aisy.202000070

3. Jayatunga MKP, Xie W, Ruder L, Schulze U, Meier C. AI in small-molecule drug discovery: a coming wave? Nature Reviews Drug Discovery. Published online February 7, 2022. doi:https://doi.org/10.1038/d41573-022-00025-1