Designing new drugs faster and more precisely, including for rare or complex diseases. This is the goal of an international research team from the Universities of Pisa and Bonn, which has developed an innovative approach to generate novel chemical compounds using artificial intelligence. The study, published in the European Journal of Medicinal Chemistry, focuses on so-called chemical language models—AI models inspired by those used in chatbots like ChatGPT—capable of reading and writing in the language of molecules.
"The aim," explains Professor Tiziano Tuccinardi from the Department of Pharmacy at the University of Pisa, "is to go beyond the limitations of traditional drug design techniques by automatically generating chemically correct, structurally original, and potentially bioactive molecules from fragments."
The researchers trained three AI models to "translate" chemical fragments—core structures, substituents, or combinations of both—into new molecules using large datasets of bioactive compounds.
"This research marks a qualitative leap in the application of AI to chemistry and pharmacology," Tuccinardi continues, "paving the way for automated and intelligent generation of molecules, with potential impact on healthcare, industry, and scientific research. It’s not just about accelerating processes, but about imagining molecular structures that the human mind might struggle to conceive."
"In line with open science principles," Tuccinardi concludes, "the source code and datasets used in the study have been made publicly available for the benefit of the scientific community. Most importantly, the project marks a significant milestone: from today, it is also possible at the University of Pisa to automatically generate new bioactive molecules—taking a concrete step toward faster, more innovative, and more accessible molecular design."
The research also involved Lisa Piazza, a PhD student in Drug and Bioactive Substance Sciences at the University of Pisa and a member of Professor Tuccinardi’s Computational Chemistry group.