Inference of genomic landscapes using ordered Hidden Markov Models with emission densities (oHMMed)

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genomes exhibit inherent inhomogeneity, with features like base composition and gene density varying along chromosomes.
• Sequential genomic data display autocorrelation, necessitating specialized analytical methods.
• Existing methods may not fully capture continuous variation patterns within genomic landscapes.

Purpose of the Study:

• To develop and present a novel class of Hidden Markov Models (HMMs) called oHMMed (ordered HMM with emission densities).
• To enable the identification and characterization of homogeneous regions within autocorrelated genomic sequences.
• To provide a biologically assumption-free approach for analyzing genomic landscapes.

Main Methods:

• Developed oHMMed, a Hidden Markov Model framework incorporating ordered emission densities.
• Modeled observed genomic data (e.g., GC content, gene number) using state-specific continuous probability distributions.
• Applied algorithms to human, mouse, and fruit fly genomes, and analyzed chromatin accessibility and epigenetic markers on human chromosome 1.

Main Results:

• Successfully partitioned genomes into regions with statistically distinguishable feature averages (GC content, gene number).
• Characterized continuous patterns of variation within genomic sequences.
• Demonstrated oHMMed's ability to analyze chromatin accessibility and epigenetic marker variations, distinguishing it from compositional domain theories.

Conclusions:

• oHMMed offers a biologically assumption-free method for characterizing genomic landscapes shaped by continuous, autocorrelated variation.
• The resulting genome segmentation facilitates the extraction of compositionally distinct regions for downstream analyses.
• This approach enhances the understanding of genomic heterogeneity and its underlying evolutionary processes.