biomvRhsmm: genomic segmentation with hidden semi-Markov model.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput technologies like next-generation sequencing (NGS) generate vast amounts of genomic data.
• Analyzing this data for genomic features like transcripts, copy number variation, and epigenetic modifications presents significant computational challenges.
• Existing methods often struggle with the complexity and volume of data from multiple genomic profiles.

Purpose of the Study:

• To develop a versatile hidden semi-Markov model (HSMM) for robust genomic segmentation.
• To create a general-purpose segmentation engine capable of handling diverse genomic data types and distributions.
• To enhance genome annotation by improving the detection of transcripts, regulatory regions, and copy number variations.

Main Methods:

• Implementation and tailoring of a hidden semi-Markov model (HSMM).
• Incorporation of genomic positions into the HSMM sojourn distribution.
• Optional prior learning using existing annotation or previous study data.
• Support for various data distributions to accommodate different genomic data types.

Main Results:

• The proposed HSMM demonstrates effective segmentation of multiple genomic profiles.
• The model provides more biologically sensible outputs by integrating genomic position information.
• Simulation benchmarking shows the HSMM achieves comparable or superior sensitivity and specificity compared to state-of-the-art segmentation models.
• The efficient implementation facilitates improved detection of genomic features.

Conclusions:

• The developed HSMM offers a powerful and flexible tool for genomic data analysis.
• This model advances genome annotation by accurately identifying various genomic elements.
• The HSMM serves as a general segmentation engine applicable to diverse high-throughput genomic datasets.