Large language models, such as those powering ChatGPT, have transformed natural language processing and the analysis of complex sequential data. In biology, protein sequences can be viewed as a language, opening new perspectives for functional annotation.
The ECxit project (Exiting the EC Classification for Better Enzyme Annotation by Deep Learning), an Inria Exploratory Action led by François Coste, focuses on enzyme function annotation. By moving beyond the traditional EC classification, it aims to develop a novel deep learning-based annotation framework built on a redesigned hierarchical classification of enzymatic functions, enabling accurate predictions directly from amino acid sequences and ultimately improving genome annotation.
The project is hosted within the Machine Learning axis of the new Bioinformatics research team BioGraphs (formerly Dyliss) at the Inria Centre at Rennes University and the IRISA research laboratory. It benefits from the Genouest platform and from close collaborations with research groups in bioinformatics, biology, and health sciences.
