FragLlama: Next-fragment prediction for molecular design.

A preprint abstract from biorxiv.org discusses the development of FragLlama, a large language model trained for molecular design. FragLlama uses custom tokens to represent molecular fragments and functional groups, and it is designed to generate molecules given one or two fragments. The model has been applied to various drug design scenarios, such as hit-to-lead and lead optimization stage drug design, PROTAC linker design, and mapping to commonly used drug design strategies. FragLlama has shown proficiency in reproducing expert-level designs and exploring novel chemical spaces, potentially enhancing the capabilities of medicinal chemists in drug design. Please note that this preprint has not undergone peer review. [Extracted from the article]