Learning Protein Structure-Function Relationships through Knowledge-guided Representation Decomposition

Proteins encode diverse functions within complex three-dimensional structures, yet most deep learning representations remain highly entangled, obscuring the biophysical signals that underlie function. Here we introduce ProtDiS, a knowledge-guided framework that decomposes pretrained protein micro-environment embeddings into biologically grounded and task-relevant dimensions. Inspired by the information bottleneck principle, ProtDiS learns representations that balance informativeness and compression, yielding structural features that are more specific, independent, and information-efficient, and achieving consistent improvements across twelve downstream tasks, with the largest gains under structure-based splits. Protein- and residue-level analyses further show that ProtDiS differentiates proteins with similar folds but divergent functions and captures fine-grained biophysical signals critical. These findings suggest that knowledge–guided decomposition provides a general and interpretable approach for structuring latent spaces in protein structural modeling. The source code and implementation details are publicly available at https://github.com/AI-HPC-Research-Team/ProtDiS.

Proteins are the essential machines of life, and their 3D shapes dictate how they function. While Artificial Intelligence (AI) is increasingly used to analyze these structures, current AI models tend to jumble the data together. This makes it difficult for scientists to understand the specific physical and biological rules driving the AI's predictions. To address this, we developed ProtDiS, a new framework that untangles this complex AI data into clear, distinct physical properties—such as a protein's shape, stability, or flexibility. By separating these features and filtering out irrelevant noise, ProtDiS makes the AI's inner workings transparent and easier for humans to interpret. When tested across twelve different biological tasks, ProtDiS consistently improved prediction accuracy. It is especially effective at distinguishing between proteins that look structurally identical but perform completely different functions. Ultimately, our tool provides a more accurate, understandable, and reliable way to use AI for protein research, which is vital for future discoveries in biotechnology and medicine.