Gene classification using codon usage and support vector machines.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Traditional gene classification methods often rely on sequence homology, which can be limited by sequence length and variance.
• Human Leukocyte Antigen (HLA) genes play crucial roles in the immune system and require accurate classification for research and clinical applications.

Purpose of the Study:

• To propose and validate a novel gene classification approach utilizing codon usage bias as input for Support Vector Machines (SVM).
• To demonstrate the efficacy of this method in classifying Human Leukocyte Antigen (HLA) sequences, addressing limitations of homology-based techniques.

Main Methods:

• DNA sequences are converted into 59-dimensional feature vectors based on relative synonymous codon usage frequencies.
• Support Vector Machines (SVM) are employed for both binary (HLA-I vs. HLA-II) and multi-class classification of HLA sequences.
• The method is tested on 1,841 Human Leukocyte Antigen (HLA) sequences.

Main Results:

• The proposed method achieves high accuracy rates: 99.3% for major HLA class classification (binary SVM).
• Multi-class SVM demonstrates excellent performance for sub-classifications: 99.73% for HLA-I and 98.38% for HLA-II.
• The classification results based on codon usage bias align with the known molecular structures and biological functions of HLA molecules.

Conclusions:

• Codon usage bias provides a robust feature set for gene classification, independent of sequence length.
• The SVM-based approach offers a highly accurate and reliable method for classifying HLA genes and their subtypes.
• This novel method enhances our ability to understand gene function and evolution through precise classification.