PLPD: reliable protein localization prediction from imbalanced and overlapped datasets.

Area of Science:

• Bioinformatics
• Computational Biology
• Proteomics

Background:

• Protein subcellular localization is crucial for understanding protein function.
• Accurate prediction methods are needed for large-scale genome analysis.
• Existing methods struggle with multi-label and imbalanced protein localization datasets.

Purpose of the Study:

• To develop a novel computational method for predicting protein subcellular localization.
• To address the challenges of multi-class, multi-label, and imbalanced datasets in protein localization prediction.
• To improve the reliability and accuracy of protein localization predictions.

Main Methods:

• A protein localization predictor based on D-SVDD (PLPD) was developed.
• The method was trained on 5184 classified proteins.
• Three new evaluation metrics were introduced for precise assessment.

Main Results:

• The PLPD method demonstrated improved accuracy in predicting protein subcellular localization.
• It effectively handles the complexities of multi-label and imbalanced datasets.
• PLPD offers a complementary approach to existing prediction methods.

Conclusions:

• The developed PLPD method provides a more reliable approach for protein subcellular localization prediction.
• This advancement aids in large-scale genome analysis and understanding protein functions.
• PLPD successfully predicted localizations for proteins with unclear experimental data.