CNN-Pred: Prediction of single-stranded and double-stranded DNA-binding protein using convolutional neural networks

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• DNA-binding proteins are crucial for essential biological processes like DNA replication, repair, and packaging.
• Classifying proteins as single-stranded DNA-binding (SSB) or double-stranded DNA-binding (DSB) is key to understanding their function.
• Current methods for predicting DSB and SSB proteins have limited accuracy.

Purpose of the Study:

• To develop a novel computational method, CNN-Pred, for accurate prediction of DSB and SSB proteins.
• To improve upon existing prediction accuracies for DNA-binding protein classification.

Main Methods:

• Feature extraction using Position Specific Scoring Matrix (PSSM) to generate mono-gram and bi-gram profiles.
• Application of a 1D-convolutional neural network (CNN) classifier to the extracted evolutionary features.
• Validation using an independent test set to evaluate prediction performance.

Main Results:

• CNN-Pred achieved enhanced prediction accuracies for DSB and SSB proteins.
• The method demonstrated an improvement of over 4% in prediction accuracy on an independent test set compared to prior studies.
• The developed tool and source code are publicly available.

Conclusions:

• CNN-Pred offers a significant advancement in the accurate prediction of DNA-binding protein types (DSB and SSB).
• The integration of PSSM features with CNNs provides a powerful approach for protein function prediction.
• The public availability of CNN-Pred facilitates further research and application in bioinformatics.