Stage Intelligence Artificielle (IA)

RNA molecules play a central role in many biological processes, and their functions are tightly linked to their 3D structures. While protein structure prediction has made significant progress thanks to models such as AlphaFold, RNA structure prediction remains an open challenge. This is largely due to the limited availability of experimental RNA structures and the high conformational flexibility of RNA.

Recent advances in multimodal foundation models (e.g., AlphaFold 3, Chai-1, Boltz-1) show that the same neural architectures can handle multiple types of biomolecules. This opens new opportunities to leverage similarities between protein and RNA folding to address data scarcity in RNA research.

At the IBISC and DSIMB laboratories, in collaboration with the Jean Zay supercomputing center, we are developing LoRAFold, a novel RNA foundation model that adapts protein-based models to RNA using parameter-efficient fine-tuning strategies.

Internship Objectives

The intern will contribute to the development and evaluation of LoRAFold. Possible tasks include:

• Exploring RNA sequence embeddings using language models.
• Adapting protein folding models to RNA through Low-Rank Adaptation (LoRA).
• Implementing structure prediction modules based on SE(3) Transformers and recycling mechanisms.
• Training and benchmarking the model on RNA3DB, a curated dataset of RNA 3D structures.
• Investigating applications in RNA-based drug discovery, with a focus on biomarkers for sepsis.

Do you have a Master's degree?, * Background in bioinformatics, computational biology, computer science, or applied mathematics.

• Experience with deep learning frameworks (PyTorch, TensorFlow, or equivalent).
• Familiarity with transformer architectures and/or geometric deep learning is a plus.
• Interest in structural biology and RNA biology.

Practical Information