Development and validation of multiple machine learning algorithms for the classification of G-protein-coupled...

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• G-protein-coupled receptors (GPCRs) are crucial drug targets, necessitating accurate function prediction.
• Existing methods for GPCR classification struggle with feature redundancy and extraction efficiency.
• Incremental large-scale GPCR datasets pose challenges for traditional machine learning models.

Purpose of the Study:

• To develop an efficient machine learning model for GPCR function prediction.
• To address the feature synonym problem in GPCR classification.
• To create a novel feature extraction strategy for enhanced GPCR analysis.

Main Methods:

• A new method based on functional word clustering and integration was developed.
• Features were clustered into synonym groups using molecular substitution matrices and transition scores.
• A feature knowledge base was created, and GPCR sequences were transformed into feature vectors.
• Machine learning models (NB, RF, SVM, MLP) were trained and tested on large GPCR datasets.

Main Results:

• The novel feature extraction strategy significantly improved GPCR classification performance.
• Models achieved high accuracy across multiple evaluation criteria compared to state-of-the-art methods.
• The approach demonstrated effectiveness on datasets of 8354 and 12,731 GPCRs.

Conclusions:

• The proposed feature extraction method offers an effective solution for the feature synonym problem in GPCR classification.
• This work provides a robust theoretical design for hierarchical GPCR classification.
• The study highlights the potential of machine learning in advancing GPCR research.