iPNHOT: a knowledge-based approach for identifying protein-nucleic acid interaction hot spots

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Protein-nucleic acid interactions are crucial for gene regulation and expression.
• Hot spots are key residues that determine binding affinity in these interactions.
• Identifying hot spots in protein-nucleic acid interfaces is less studied than in protein-protein interactions due to limited mutagenesis data.

Purpose of the Study:

• To develop a novel computational model, iPNHOT, for predicting hot spot residues in protein-nucleic acid interfaces.
• To improve the accuracy and efficiency of identifying critical residues involved in protein-DNA and protein-RNA binding.

Main Methods:

• Generated 97 sequential and structural features for modeling.
• Employed a two-step feature selection strategy.
• Utilized a Support Vector Machine (SVM) algorithm, with the final model based on 7 selected features, including two novel features (∆SASsa^1/2 and esp3).

Main Results:

• The iPNHOT model achieved an F1 score of 0.725 and an AUROC of 0.807 on a subset from ProNIT.
• These scores significantly outperform the state-of-the-art model mCSM-NA (F1: 0.407, AUROC: 0.670).
• Independent testing confirmed iPNHOT's superior performance compared to other existing methods.

Conclusions:

• A new computational model, iPNHOT, was developed to predict hot spots on protein-DNA and protein-RNA interfaces.
• iPNHOT demonstrates superior predictive performance over current state-of-the-art models.
• The iPNHOT model is accessible to researchers via a webserver for broader application.