Description of the thesis topic

Deep generative model of active molecules conditioned by cellular phenotype.
The PhD student will contribute to the development of innovative artificial intelligence approaches for the discovery of new therapeutic chemical compounds. The project involves designing a generative model conditioned by cellular phenotypic cell, with the goal of creating molecules capable of reproducing a desired cellular state. The candidate will take part in defining scientific objectives, setting up evaluation protocols, and promoting results through publications and interdisciplinary collaborations.
The work will include the design, training, and evaluation of diffusion models applied to molecular graphs, as well as the development of robust metrics to characterize the quality and diversity of the generated compounds. The candidate will handle large-scale datasets combining chemical structures and cell phenotypes, and implement advanced deep learning techniques. Regular exchanges with an industrial partner will provide the opportunity to explore the synthesis and experimental testing of identified candidates.

Work Context

The thesis will take place in a research environment at the interface of biology, chemistry, and artificial intelligence. The project is led by a multidisciplinary team specialized in cell image analysis and multimodal modeling, providing a stimulating framework for scientific innovation. The candidate will benefit from close supervision, access to state-of-the-art computing infrastructure, and opportunities to collaborate with both academic and industrial researchers, within a dynamic aimed at the discovery of first-in-class drugs.

Constraints and risks

Work on screen