Scientists Struggle with Precision of AI Technology in Developing Novel Drug Candidates

The UNC School of Medicine lab, led by Bryan Roth, MD, PhD, in collaboration with researchers from UCSF, Stanford, and Harvard, has made a groundbreaking discovery in the field of drug discovery. Their study has shown that a protein prediction technology, AlphaFold2, can accurately predict protein structures and identify potential drug candidates for the treatment of various conditions, particularly neuropsychiatric disorders.

In a recent paper published in Science, the researchers demonstrated that AlphaFold2 can accurately model ligand binding sites on proteins, where drugs attach to initiate therapeutic effects. This finding is crucial for the development of new and effective drugs for conditions like Alzheimer’s disease and schizophrenia.

The study involved testing AlphaFold2’s accuracy by comparing its predictions with known protein structures in a retrospective study and using it to predict interactions with unknown proteins in a prospective study. The results were promising, with successful hit rates of around 50% for the sigma-2 receptor and 20% for the 5-HT2A receptor.

Roth and his colleagues selected these proteins specifically because AlphaFold2 had no prior information about them, making it a true test of the technology’s capabilities. The researchers were able to identify potential drug candidates for these proteins, opening up new possibilities for drug discovery in the future.

“This work would be impossible without collaborations among several leading experts at UCSF, Stanford, Harvard, and UNC-Chapel Hill,” said Roth. The team plans to further explore the applications of AlphaFold2 in drug discovery for other therapeutic targets and target classes.

Overall, this study represents a significant advancement in the field of drug discovery, showcasing the potential of artificial intelligence in revolutionizing the way we develop new treatments for various medical conditions.