TSNAPred: predicting type-specific nucleic acid binding residues via an ensemble approach

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Protein-nucleic acid interactions are crucial for biological processes.
• Existing methods for identifying these interactions are often time-consuming or lack specificity.
• Accurate identification of binding residues is key to understanding protein function.

Purpose of the Study:

• To develop an accurate and efficient computational method for identifying nucleic acid-binding residues.
• To differentiate between DNA-binding and RNA-binding residues.
• To improve upon existing state-of-the-art methods in prediction performance.

Main Methods:

• Developed TSNAPred, an ensemble predictor combining LightGBM and capsule network.
• Utilized features derived from protein sequences.
• Employed a sliding window technique for long-distance dependency extraction and a weighted ensemble strategy.

Main Results:

• TSNAPred effectively identifies type-specific nucleic acid binding residues.
• Achieved a 5% to 10% improvement in AUC value for discriminating DNA-binding and RNA-binding residues compared to other methods.
• Demonstrated robust performance on a test set.

Conclusions:

• TSNAPred offers an effective computational solution for identifying nucleic acid-binding residues.
• The method enhances the understanding of protein-nucleic acid interactions.
• Provides a valuable tool for bioinformatics research with available code and dataset.