INFUSSE is a breakthrough framework that integrates protein sequence embeddings from Large Language Models with structural information through graph-based deep learning to predict protein properties at the single-residue level.

• Combines the power of LLM-generated embeddings with graph-based structural representations
• Uses a diffusive Graph Convolutional Network (diff-GCN) for information integration
• Enhances prediction capabilities for crucial single-residue properties
• Demonstrates how AI can unlock deeper understanding of protein behavior

This research advances the field of molecular biology by providing a more holistic approach to protein analysis, potentially accelerating drug discovery and protein engineering applications.

Integrating protein sequence embeddings with structure via graph-based deep learning for the prediction of single-residue properties