Gene finding with a hidden Markov model of genome structure and evolution

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Genome sequencing is rapidly expanding, revealing diverse evolutionary patterns across functional regions.
• Exploiting these evolutionary differences can enhance genomic annotation.
• Existing comparative gene finders have limitations in modeling evolutionary processes.

Purpose of the Study:

• To design a novel probabilistic model for genome structure and evolution.
• To improve the accuracy of gene finding through enhanced evolutionary modeling.

Main Methods:

• Developed an Evolutionary Hidden Markov Model (EHMM) integrating Hidden Markov Models (HMMs) with phylogenetic tree-based evolutionary models.
• Estimated all model parameters, including the phylogenetic tree, using maximum likelihood.
• Algorithms exhibit linear time complexity concerning alignment length and genome number.

Main Results:

• The EHMM effectively models evolutionary correlations using phylogenetic trees for any number of aligned genomes.
• Demonstrated the utility of modeling sequence evolution for gene finding.
• Validated the model's performance through simulations and analysis of human/mouse gene pairs.

Conclusions:

• The EHMM provides a robust framework for genomic annotation by incorporating evolutionary dynamics.
• This model offers a significant advancement over existing comparative gene finding methods.
• The software is freely available online for broader research application.