Using hidden Markov models and observed evolution to annotate viral genomes.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Single-stranded RNA (ssRNA) viruses possess compact genomes, often using overlapping reading frames to maximize coding capacity.
• Overlapping genes present evolutionary constraints, where mutations can affect multiple protein-coding regions.
• Compositional biases in overlapping regions have been previously used heuristically for gene detection.

Purpose of the Study:

• To develop a probabilistic Hidden Markov Model (HMM) for accurate annotation of ssRNA viral genomes, focusing on overlapping regions.
• To investigate the utility of differentiated nucleotide compositional patterns within an HMM framework for gene prediction.
• To extend the HMM approach to a Phylogenetic HMM (Phylo-HMM) by incorporating evolutionary information for improved annotation.

Main Methods:

• Development of a single sequence annotation (SSA) procedure using an HMM framework to analyze nucleotide composition in overlapping regions.
• Parameterization of the HMM for annotation within a missing data framework.
• Extension of the HMM to a Phylo-HMM by integrating aligned sequence data (14 HIV2 strains) to leverage evolutionary information.

Main Results:

• The SSA method demonstrated encouraging results in annotating ssRNA viral genomes, including 14 strains of HIV2.
• The Phylo-HMM extension, applied to aligned HIV2 sequences, showed performance improvements compared to the single-sequence approach.
• The study illustrates the potential of Phylo-HMM for enhancing ssRNA viral genome annotation by incorporating evolutionary context.

Conclusions:

• The HMM-based approach effectively utilizes compositional patterns for ssRNA viral genome annotation.
• Incorporating evolutionary information through Phylo-HMM significantly improves annotation accuracy.
• This framework offers a powerful tool for understanding the complex coding strategies of ssRNA viruses.