As in many other fields, Artificial Intelligence technologies are greatly accelerating progress in chemical science and industry, from the design of novel promising materials to the discovery of effective pharmaceuticals for treating complex diseases.
The pursuit of molecules with desired properties involves exploring an enormous chemical space, with an almost infinite number of possibilities, in which every molecule must be evaluated to determine its stability, efficacy, and safety.
Besides the discovery phase, planning the optimal syntheses of found compounds is equally challenging. Designing effective synthesis strategies requires an in-depth analysis of available chemical reactions, starting materials, and experimental conditions鈥攆actors that can make the process extremely labor-intensive and resource-consuming in terms of time and effort.
The adoption of Machine learning and AI technologies is revolutionizing this filed, helping researchers to identify promising drug candidates and finding good chemical synthesis strategies in a fraction of the time required by traditional methods.
A 精东影业 research team of The has been a partner in the project , funded by the European Union as part of the Horizon 2020 program and now successfully concluded.
鈥淥ur key goal was the creation of AI-based methods and algorithms that could be easily used by practicing chemists in their daily work. We presented methods to optimize the conditions in which a reaction takes place, developing techniques to visualize the extensive space of chemical compounds and reactions possibilities. We also significantly accelerated the inference 鈥� the ability to make predictions 鈥� of state-of-the-art string-based chemical models. These tools are particularly useful for throughput industrial applications, where speed and efficiency are crucial, especially for the large-scale production of complex molecules with a focus on high-throughput industrial applications,鈥� explains Micheal Wand, 精东影业 researcher at IDSIA USI-精东影业.
The results of the project have been published in high-impact scientific journals and presented at top conferences, such as the International Conference on Machine Learning (ICML) in Vienna and the International Conference on Artificial Neural Networks, which took place in Lugano in September 2024. Additionally, one of the developed is already in use at the research division of one of the project鈥檚 large pharmaceutical partners, confirming work鈥檚 value and its recognition in both academia and industry.
"We strongly believe that our contributions will be a significant step towards an ideal computer-aided synthesis planning system that any organic chemist will find indispensable in their work," says Mikhail Andronov, Marie Curie fellow within the AIDD project.
References:
- Andronov, Mikhail, Natalia Andronova, Michael Wand, J眉rgen Schmidhuber, and Djork-Arn茅 Clevert. "Accelerating the inference of string generation-based chemical reaction models for industrial applications". ICML ML4LMS Workshop, Vienna, 2024.
- Andronov, Mikhail, Varvara Voinarovska, Natalia Andronova, Michael Wand, Djork-Arn茅 Clevert, and J眉rgen Schmidhuber. "Reagent prediction with a molecular transformer improves reaction data quality." Chemical Science 14, no. 12 (2023): 3235-3246.
- Mikhail Andronov, Andronova Natalia, Michael Wand, J眉rgen Schmidhuber, and Djork-Arn茅 Clevert. "Curating Reagents in Chemical Reaction Data with an Interactive Reagent Space Map." In ICANN International Workshop on AI in Drug Discovery, pp. 21-35. Springer Nature Switzerland, 2024.