DeepMC-iNABP: Deep learning for multiclass identification and classification of nucleic acid-binding proteins

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Nucleic acid-binding proteins (NABPs), encompassing DNA-binding proteins (DBPs) and RNA-binding proteins (RBPs), are essential regulators of gene expression.
• Current identification methods face challenges, including overlooking dual DNA- and RNA-binding proteins (DRBPs) and inaccuracies in distinguishing between DBPs and RBPs (cross-prediction).

Purpose of the Study:

• To develop a computational predictor, DeepMC-iNABP, to accurately identify NABPs.
• To address the limitations of existing methods by effectively identifying DRBPs and mitigating cross-prediction issues.

Main Methods:

• A multiclass classification strategy was employed using deep learning approaches.
• The DeepMC-iNABP model was trained on four distinct data classes: DBPs, RBPs, DRBPs, and non-NABPs.

Main Results:

• DeepMC-iNABP demonstrated a significant advantage in identifying DRBPs.
• The model showed an ability to alleviate the cross-prediction problem between DBP and RBP predictors.
• Performance was validated on in-house and two independent test datasets.

Conclusions:

• DeepMC-iNABP offers an effective solution for the accurate identification of various classes of nucleic acid-binding proteins.
• The developed tool enhances the study of gene expression regulation by improving the prediction of NABPs, particularly DRBPs.