Resolving complex tandem repeats with long reads.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Resolving tandem repeats (TRs), also known as microsatellites, is crucial for understanding the human genome and identifying genetic disorders.
• Existing short-read sequencing tools have limitations in resolving diverse TRs, while single-molecule sequencing offers potential but faces challenges due to high error rates.
• Accurate identification and estimation of TRs in long reads are essential for advancing genome assembly, alignment, and variation calling.

Purpose of the Study:

• To develop a novel reference-based probabilistic approach, PacmonSTR, for identifying TR regions and estimating TR elements in long DNA reads.
• To address the challenges of high error rates in single-molecule sequencing data for TR analysis.

Main Methods:

• PacmonSTR employs a multistep approach using a reference region and TR element as input.
• TR regions are identified using a 3-stage modified Smith-Waterman algorithm on long DNA reads.
• TR element counts are estimated with a pair-Hidden Markov Models-based method, followed by genotype selection using Gaussian mixture models.

Main Results:

• PacmonSTR successfully identifies TR regions and estimates the number of TR elements within long DNA reads.
• The method provides TR-based genotype selection (homozygous/heterozygous) by leveraging Akaike information criteria and coverage expectations.
• This approach enhances the ability to resolve TRs in the context of high-error, long-read sequencing data.

Conclusions:

• PacmonSTR offers a robust solution for accurate TR identification and quantification in long-read sequencing data.
• The developed method improves upon existing techniques by effectively handling the error profiles inherent in single-molecule sequencing.
• This advancement facilitates more comprehensive genomic analysis, particularly in regions with complex tandem repeats.