NCBS Develops AI Tool to Decode Disordered Protein Interactions

Researchers at the National Centre for Biological Sciences (NCBS), Tata Institute of Fundamental Research (TIFR), Bengaluru, have developed a deep-learning tool called Disobind to predict how intrinsically disordered proteins (IDPs) bind to their partner proteins.

Background
Traditionally, protein function has been understood as being dependent on a stable, well-defined three-dimensional structure. However, intrinsically disordered proteins challenge this view as they do not adopt a fixed structure and instead remain flexible and shapeshifting.

About Intrinsically Disordered Proteins (IDPs)
IDPs are central to life and play key roles in:

• Cellular signalling networks
• Helping proteins move and locate binding partners
• Gene regulation (switching genes on or off)
• Protein folding and quality control
• Formation of dynamic cellular hubs known as biomolecular condensates

About the Disobind Tool
Disobind analyses protein sequences to predict which regions of a flexible, disordered protein are likely to interact with another protein. It uses protein language models, a form of artificial intelligence trained on millions of known protein sequences.

Key Findings

• Disobind showed consistently higher accuracy when tested on new protein pairs that were not part of its training data.
• Its predictive performance improved further when combined with AlphaFold-Multimer outputs.

About AlphaFold
AlphaFold, developed by DeepMind, is an AI system that predicts a protein’s three-dimensional structure. AlphaFold-Multimer extends this capability to protein–protein interactions.