Polymorphisms predicting phylogeny in hepatitis B virus.

Area of Science:

• Virology
• Genetics
• Bioinformatics

Background:

• Hepatitis B virus (HBV) has ten genotypes (A-J) influencing disease transmission, progression, and treatment efficacy.
• Accurate HBV genotyping is crucial but often limited by accessibility and cost of sequencing.
• Identifying informative genetic markers can simplify and improve HBV genotyping methods.

Purpose of the Study:

• To identify specific amino acid (aa) sites within the HBV genome that are most informative for determining HBV genotype.
• To utilize a machine learning approach, specifically random forest algorithms (RFA), for genotype prediction.
• To assess the potential of these informative sites for improving HBV genotyping accessibility.

Main Methods:

• Downloaded 5,496 genome-length HBV sequences from a public database, excluding recombinant sequences and specific genotypes (I, J).
• Translated HBV genes into amino acid sequences and concatenated them.
• Applied RFA to identify aa sites predictive of genotype and assessed site covariation using mutual information.

Main Results:

• Ten informative amino acid sites were identified, enabling confident discrimination between HBV genotypes A-H.
• Five of these sites were located in the Polymerase protein (4 in the spacer domain, 1 in reverse transcriptase), four in the Surface protein, and one in HBx.
• No informative sites were found in the Core protein; amino acid properties at informative sites were generally not conserved across genotypes.

Conclusions:

• Random forest algorithm analysis is a powerful tool for identifying amino acid sites that predict HBV lineage.
• The identification of numerous informative sites in the Polymerase spacer domain challenges conventional views of its importance.
• These findings enhance the ease of HBV genotype prediction from limited sequence data and may inform future HBV evolution studies.