Generator based approach to analyze mutations in genomic datasets.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Conventional genomic sequence comparison relies on sequence alignment tools, which can be computationally intensive and less effective for large datasets.
• Existing methods may struggle with capturing the statistical properties and evolutionary dynamics of genomic sequences efficiently.

Purpose of the Study:

• To propose a novel computational approach for comparing genomic sequences based on learned sequence generators.
• To enable efficient statistical distance computation, clustering, and evolutionary characterization of large genomic datasets.
• To identify mutation activity regions within genomic sequences without direct alignment.

Main Methods:

• Developed a data-driven approach to learn state machine-based sequence generators for genomic sequences.
• Utilized these generators to compute statistical distances between sequences, independent of sequence length.
• Applied the method to cluster genomic datasets, analyze evolutionary trajectories, and detect mutation hotspots.

Main Results:

• Demonstrated an efficient method for computing statistical distances between large sets of genomic sequences.
• Successfully clustered diverse strains of SARS-CoV-2 viral sequences, revealing their evolutionary patterns.
• Identified specific regions within SARS-CoV-2 sequences exhibiting mutation activity, aiding potential alignment techniques.

Conclusions:

• The proposed sequence generator comparison method offers a fast and efficient alternative to traditional alignment for genomic data analysis.
• This technique provides valuable insights into genomic sequence evolution, clustering, and mutation detection.
• The approach has practical applications in analyzing viral evolution, such as with SARS-CoV-2 strains.