INAB: identify nucleic acid binding domain via cross-modal protein language models and multiscale computation

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Protein-nucleic acid interactions are vital for gene regulation and editing.
• Experimental methods for identifying binding domains are costly and slow.
• Computational tools are needed to accelerate the study of these interactions.

Purpose of the Study:

• To introduce a novel computational framework for predicting nucleic acid-binding domains in proteins.
• To leverage cross-modal protein language models and a multiscale architecture for enhanced prediction accuracy.

Main Methods:

• Developed a framework integrating cross-modal protein language models.
• Utilized a structurally annotated benchmark dataset with hierarchical, proximity-based labels.
• Employed a multiscale computational architecture for binding likelihood quantification.

Main Results:

• Achieved state-of-the-art performance in nucleic acid-binding domain prediction.
• Demonstrated the effectiveness of the multimodal learning approach.
• Provided a valuable open resource for researchers.

Conclusions:

• The proposed framework offers a powerful computational solution for identifying nucleic acid-binding domains.
• This work advances multimodal learning in protein-nucleic acid interaction analysis.
• The resource will accelerate discoveries in functional genomics and drug design.