AIM-GPT (label PRIME CNRS)

Artifical Inteligence for Material Generation & Predictive Tools

Label PRIME de la Mission pour les initiatives transverses et interdisciplinaires (MITI) du CNRS.

The collaboration between the Lens Computer Science Research Centre (CRIL) and the Catalysis and Solid State Chemistry Unit (UCCS) since 2020 on the use of artificial intelligence (AI) for chemistry and materials illustrates a fruitful and sustainable synergy. This interdisciplinary collaboration combines AI skills with in-depth knowledge of materials and chemistry, promoting an innovative approach to the generation of materials and the prediction of their properties. It has resulted in several PhD theses funded by ANR VIVAH, such as those by Gokhan Tahil and Astrid Klipfel, and ongoing research projects, such as those of Elohan Veillon and Tom Ternisien, demonstrating a long-term commitment to training young researchers. The researchers have published several articles in prestigious journals, demonstrating the quality and impact of their work, notably in journals such as the Journal of Chemical Information and Modelling and at international conferences such as AAAI and IJCAI, highlighting the international recognition of their research. The AIM-GPT collaboration strengthens the position of laboratories in the French research landscape, particularly in Hauts-de-France, by promoting innovation and responding to regional priorities such as energy.

By integrating AI into materials chemistry, researchers are opening up new avenues of research, thereby influencing practices and methodologies in these fields and increasing the visibility of laboratories at the local, national and international levels. The work aims to solve complex problems such as the generation of new materials for specific applications, which can have repercussions on industry and the environment, thereby strengthening the societal impact of research. In summary, the AIM-GPT collaboration between CRIL and UCCS is an example of successful interdisciplinary research, which not only enriches scientific knowledge but also contributes significantly to the laboratories’ influence at the local, national and international levels.

Defended theses (ANR VIVAH funding)

• Gokhan Tahil encadré par Daniel Le Berre et Sébastien Tilloy
• Astrid Klipfel encadré par Yael Frégier, Adlane Sayede et Zied Bouraoui

Ongoing PhD theses

• Elohan Veillon (DGF Artois) encadré par Adlane Sayede et Zied Bouraoui
• Tom Ternisien (PIA4 MAIA) encadré par Florence Pilard (LG2A), Sébastien Tilloy et Nejat Arinik Vers la Synthèse de Nouveaux Dérivés de Cyclodextrines Assistée par l’Intelligence Artificielle – (SYNTIA)

Ongoing collaboration

• Jed Alwa DUERSH PostDoc PIA4 MAIA encadré par Adlane Sayede et Zied Bouraoui

Publications

Gökhan Tahıl, Fabien Delorme, Daniel Le Berre, Éric Monflier, Adlane Sayede, Sébastien Tilloy, Stereoisomers Are Not Machine Learning’s Best Friends in Journal of Chemical Information and Modeling, vol. 64, n° 14, pp. 5451–5469, 2024.

Astrid Klipfel, Yaël Fregier, Adlane Sayede, Zied Bouraoui, Vector Field Oriented Diffusion Model for Crystal Material Generation in The 38th Annual AAAI Conference on Artificial Intelligence, vol. 38, pp. 22193-22201, 2024.

Astrid Klipfel, Yaël Frégier, Adlane Sayede, Zied Bouraoui, Optimized Crystallographic Graph Generation for Material Science in Thirty-Second International Joint Conference on Artificial Intelligence {IJCAI-23}, International Joint Conferences on Artificial Intelligence Organization, pp. 7145- 7148, 2023.

Astrid Klipfel, Yaël Frégier, Adlane Sayede, Zied Bouraoui, Unified Model for Crystalline Material Generation in 32nd International Joint Conference on Artificial Intelligence, IJCAI 2023, vol. 2023- August, pp. 6031 – 6039, 2023.

Astrid Klipfel, Zied Bouraoui, Olivier Peltre, Yaël Fregier, Najwa Harrati, Adlane Sayede, Equivariant Message Passing Neural Network for Crystal Material Discovery in AAAI Conference on Artificial Intelligence, vol. 37, pp. 14304-14311, 2023.

Gökhan Tahil, Fabien Delorme, Daniel Le Berre, Éric Monflier, Adlane Sayede, Sébastien Tilloy, Curated dataset of association constants between a cyclodextrin and a guest for machine learning in Chemical Data Collections, vol. 45, pp. 101022, 2023.