Computer-aided drug discovery: From traditional simulation methods to language models and quantum computing

ZR Pei, CELL REPORTS PHYSICAL SCIENCE, 5, 102334 (2024).

DOI: 10.1016/j.xcrp.2024.102334

Drug discovery is a central topic at the intersection of structural biology, biochemistry, and medicine, involving significant challenges like high cost (usually more than billions of dollars), low success rates (typically <10%), and extremely long cycles (often over a decade). Computer-aided drug discovery (CADD) shows huge advantages in addressing these challenges and accelerating the process, making it an indispensable tool in the pharmaceutical industry and scientific research. Here, we review the latest proceedings in this active field and explore the transformative opportunities presented by machine learning, language models, and quantum computing in CADD. The recent development of AlphaFold 2 and 3, state-of-the-art machine learning models, marks a significant advancement in CADD. AlphaFold 3 excels in accurately predicting protein structures, identifying potential docking sites, and facilitating high-throughput docking screenings, thereby streamlining the entire drug discovery process. This model represents a substantial improvement over its predecessors, offering higher accuracy and reliability in structural predictions. Beyond AlphaFold, various machine learning techniques are revolutionizing different stages of drug discovery, from virtual screening to predictive modeling of drug-target interactions. Language models, such as the GPT models, offer promising applications in automating literature reviews, generating research hypotheses, and aiding in interpreting complex biological data. Additionally, quantum computing holds the potential to solve intricate molecular simulations and optimization problems that are currently intractable for classical computers, although its practical implementation remains in the early stages.

Return to Publications page