A new supervised over-sampling algorithm with application to protein-nucleotide binding residue prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein-nucleotide interactions are vital for biological processes.
• Accurate identification of binding residues from protein sequences aids function annotation and drug design.
• Predicting protein-nucleotide binding residues presents a class imbalance challenge, with fewer binding than non-binding residues.

Purpose of the Study:

• To address the negative impact of class imbalance on protein-nucleotide binding residue prediction.
• To propose a novel supervised over-sampling algorithm to synthesize minority class samples.
• To develop an effective predictor for protein-nucleotide binding residue identification.

Main Methods:

• Developed a supervised over-sampling algorithm to generate synthetic minority class samples.
• Implemented a predictor named TargetSOS based on the proposed over-sampling technique.
• Evaluated the predictor using cross-validation and independent validation tests on protein-nucleotide interaction datasets.

Main Results:

• The proposed over-sampling algorithm effectively alleviates class imbalance in protein-nucleotide binding residue prediction datasets.
• TargetSOS demonstrated improved prediction performance compared to existing methods.
• Experimental results confirmed the effectiveness of the supervised over-sampling approach.

Conclusions:

• Supervised over-sampling is a viable strategy to enhance machine learning models for imbalanced biological data.
• TargetSOS provides an effective solution for accurate protein-nucleotide binding residue prediction.
• The developed web-server and datasets are publicly available to facilitate further research.