Researchers at the Lamarr Institute and the Bonn-Aachen International Center for Information Technology (b-it) at University of Bonn have made a significant breakthrough by developing an AI-based model capable of predicting potential active substances with unique therapeutic properties. Utilizing an advanced chemical language model – essentially a “ChatGPT for molecules” – the team trained the AI to generate precise chemical structural formulas that could be highly effective in creating new medications. The study, led by Prof. Dr. Jürgen Bajorath Lamarr Area Chair for Life Sciences and professor and chair of Life Science Informatics at the b-it/University of Bonn, has been published in Cell Reports Physical Science.
Addressing Challenges in Polypharmacology with AI
Prof. Dr. Jürgen Bajorath, who heads the Interdisciplinary Research Area Life Sciences (LS Area) at the Lamarr Institute, emphasized the importance of this model for pharmaceutical research. “In the field of drug discovery, dual-target drugs are highly desirable due to their polypharmacological effects. These compounds can simultaneously impact multiple cellular processes and signaling pathways, making them especially effective against complex diseases like cancer,” he stated. However, designing these sophisticated molecules remains a significant challenge.
Trained with over 70,000 molecular pairs, the AI model was designed to generate compounds capable of binding to two different target proteins. Dual-target inhibitors, compared to combining single-target drugs, offer a more synchronized pharmacokinetic profile and minimize the risk of adverse drug interactions.
Innovative AI Approach in Drug Discovery at Lamarr
The AI model is rooted in Lamarr’s Triangular AI concept, which integrates heterogeneous life science data, diverse scientific contexts, and interdisciplinary knowledge to create tailored predictive modeling. After an initial training phase, the research team applied fine-tuning to focus the model on different enzyme and receptor classes, leading to the accurate production of dual-target compounds.
According to Prof. Bajorath, “To a certain extent, it triggers ‘out of the box’ ideas and comes up with original solutions that can lead to new design hypotheses and approaches.” This feature opens up creative possibilities in drug discovery, advancing the pharmaceutical research landscape.
Collaborative Research Effort with Real-World Impact
This study showcases the collaborative spirit of the LS Area at the Lamarr Institute, which emphasizes explainable AI (XAI) in advancing life sciences research. This project exemplifies the Institute’s dedication to combining machine learning and AI technologies with key disciplines in medical research and drug discovery. This pioneering research, conducted at the Lamarr Institute, the University of Bonn, and b-it, exemplifies the collaborative spirit and interdisciplinary excellence that drive AI advancements in life sciences, reinforcing the Lamarr Institute’s mission to strengthen Europe as a leader in AI-driven innovation.
Study Publication
Sanjana Srinivasan and Jürgen Bajorath: Generation of Dual-Target Compounds Using a Transformer Chemical Language Model; Cell Reports Physical Science; DOI: 10.1016/j.xcrp.2024.102255