BITS Pilani Faculty Co-Authors AI Tool to Identify DNA-Binding Protein Regions

The study, published in the journal Nucleic Acids Research, addresses a major challenge in identifying specific protein residues that interact with DNA, a process crucial for gene regulation, DNA replication and repair, and the organisation of genetic material, the Institute said in a release on Tuesday.

The research was led by scientists from Virginia Commonwealth University, the US, and Xinyang Normal University, China, in collaboration with Hebei University of Technology, China.

DNAreader combines a specialised batch-training strategy with contrastive learning to distinguish DNA-binding residues from those that interact with other molecules or do not bind to any molecule.

The approach is designed to reduce incorrect predictions, particularly in flexible protein regions that lack a fixed structure.

Dr Basu said identifying DNA-binding residues was especially challenging in proteins containing intrinsically disordered regions.

By integrating transformer-based learning with contrastive learning, DNAreader provides a unified framework for analysing structured and disordered regions and improves the distinction between DNA-binding and other residues.

In evaluations against 14 existing prediction tools using independent test datasets with low sequence similarity, DNAreader achieved an area under the receiver operating characteristic curve (AUROC) score of 0.86 to 0.89, compared with 0.75 to 0.79 for the leading existing tools, according to the researchers.

The team demonstrated the tool’s performance using the human estrogen-related receptor ERRβ, in which DNA-binding residues were identified across structured and disordered regions.

It also correctly avoided false predictions in an unrelated control enzyme.

DNAreader is freely accessible through a web server that accepts protein sequences in FASTA format. Its source code and underlying datasets have also been made publicly available to facilitate further research.

The study highlights the application of artificial intelligence in computational biology and could support research into protein-DNA interactions and related biological processes.