Indian Scientists Develop ‘Disobind’ to Predict Binding of Disordered Proteins

Indian Scientists Develop AI Tool to Decode Binding of Disordered Proteins

Researchers at the National Centre for Biological Sciences (NCBS), under the Tata Institute of Fundamental Research , Bengaluru, have developed a deep-learning-based tool that can predict how intrinsically disordered proteins (IDPs) interact with their binding partners. The breakthrough addresses a long-standing challenge in molecular biology and holds promise for advances in disease research and drug discovery .

Why Intrinsically Disordered Proteins Are Important

Unlike conventional proteins that fold into stable three-dimensional structures, intrinsically disordered proteins do not have a fixed shape. Despite this apparent lack of structure, they play critical roles in cellular functioning . IDPs are involved in:

• Cell signalling and regulatory networks

• Gene expression control

• Protein folding and quality control

• Formation of biomolecular condensates

Their structural flexibility allows functional versatility but has made them difficult to study using traditional structural biology tools .

Disobind: The Deep-Learning Breakthrough

The newly developed tool, named Disobind , predicts which regions of a disordered protein are likely to bind to a specific partner protein. It uses protein language models , a class of artificial intelligence trained on millions of protein sequences , to identify interaction-prone regions directly from sequence data.

A key strength of Disobind is that it:

• Does not require prior structural information

• Does not rely on sequence alignments

• Explicitly accounts for the identity of the binding partner , which is crucial for IDP interactions

Performance and Benchmarking

The research team, led by Kartik Majila , benchmarked Disobind against existing state-of-the-art predictors such as AlphaFold-Multimer and AlphaFold3 .

Disobind consistently demonstrated higher predictive accuracy , especially when tested on previously unseen protein pairs . Notably, when combined with AlphaFold-Multimer outputs, overall prediction performance improved further, indicating that Disobind complements structure-based approaches rather than replacing them .

Applications in Disease Research and Drug Design

According to Shruthi Viswanath , head of the Integrative Structural Biology Lab at NCBS, Disobind can help:

• Identify disease-associated interaction motifs

• Reveal new therapeutic intervention points

• Study protein systems involved in immune signalling, cancer, and neurodegenerative disorders

Importantly, Disobind has been released as open-source software , enabling researchers worldwide to freely apply and build upon the tool.