Deep neural learning based protein function prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein function prediction is essential for annotating uncharacterized proteins.
• Current Deep Neural Network (DNN) methods often focus on small datasets or single features, limiting large-scale, multi-feature applications.
• Effective methods for large-scale, multi-feature protein prediction require further development.

Purpose of the Study:

• To propose an advanced Deep Neural Network (DNN) based approach, termed IGP-DNN, for protein function prediction.
• To address the limitations of existing methods in handling large-scale, multi-feature protein data.
• To enhance the accuracy and efficiency of protein function annotation.

Main Methods:

• Utilized the Grasshopper Optimization Algorithm (GOA) and Intuitionistic Fuzzy c-Means clustering (IFCM) for protein function module feature extraction.
• Employed Kernel Principal Component Analysis (KPCA) for dimensionality reduction of protein attribute information.
• Integrated extracted module features and attribute features as input for a multi-hidden layer DNN.

Main Results:

• The proposed IGP-DNN approach demonstrated improved performance in protein function prediction.
• Achieved an F-measure value of 0.4436 on the DIP dataset, indicating superior predictive capability.
• Successfully integrated diverse protein features for enhanced classification accuracy.

Conclusions:

• IGP-DNN offers a robust and effective solution for large-scale, multi-feature protein function prediction.
• The integration of GOA, IFCM, and KPCA with DNN significantly enhances prediction accuracy.
• This method provides a valuable tool for advancing protein annotation and understanding biological functions.