A revolutionary AI tool created by Australian researchers may dramatically accelerate the drug discovery process, offering a cheaper and more effective method. The tool, named PSICHIC (PhySIcoCHemICal), has been designed to predict the way in which molecules and proteins interact within the human body. Its use could significantly reform the early stages of drug discovery.
PSICHIC was devised by a research group led by Monash University which has published their findings in the journal, Nature Machine Intelligence. Dr Lauren May of Monash’s Institute of Pharmaceutical Sciences (MIPS) lauds the tool’s potential to transform the drug discovery process, suggesting its application could refine virtual screening and improve understanding of protein-molecule interactions.
Developing new drugs is an intricate and costly undertaking. A primary difficulty for researchers is discerning how different molecules react with bodily proteins. This understanding is vital, as drug efficacy and safety hinge on these interactions. Presently, predicting these interactions frequently necessitates intricate 3D structures, which are expensive and lengthy to produce. PSICHIC aims to alleviate this problem by utilising AI to predict protein-molecule interactions based solely on sequence data, and eliminating the need for 3D structuring.
Professor Geoff Webb, an AI specialist from Monash’s Department of Data Science and Artificial Intelligence, highlights the expanding use of AI methodologies in enhancing drug discovery’s affordability and precision. Meanwhile, Dr Anh Nguyen from MIPS, an expert on AI approaches to drug-receptor interactions, emphasises the critical role of molecule-protein interplay in biological processes and drug creation.
Huan Yee Koh, Monash PhD candidate and lead author of the study, outlined the rationale behind developing PSICHIC for specific use in drug discovery applications. Despite its potential to improve efficiency, robustness, and cost at various points throughout the drug discovery process, Koh also noted that AI systems can suffer from too much freedom. They may memorise previously known patterns instead of learning the protein-ligand interaction mechanisms, which can inhibit the discovery of novel drugs. PSICHIC will overcome this problem.
The researchers at PSICHIC have made the data, code, and optimised model accessible to the wider scientific community. Their work could revolutionise the process of drug discovery, potentially making it quicker, cheaper, and more efficient. Through the application of this tool, we can deepen understanding of the interactions between molecules and proteins, crucial for the effectiveness and safety of new drugs. PSICHIC’s data, code and optimized model have been made available for further scientific research.
