Finding long tandem repeats in long noisy reads

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long tandem repeat expansions (>1000 nt) are linked to diseases but are difficult to study in individual genomes due to short read lengths.
• Emerging long-read sequencing technologies (≥10,000 nt) can span these repeats but suffer from high error rates (10-20%).
• Existing algorithms for tandem repeat detection are optimized for short repeats and struggle with long reads' error rates and computational demands.

Purpose of the Study:

• To develop an efficient algorithm for detecting long tandem repeat expansions in human genomes using long-read sequencing data.
• To address the challenges posed by high error rates in long reads for repetitive element detection.
• To provide a sensitive and computationally feasible method for analyzing disease-associated repeat expansions.

Main Methods:

• Developed a novel algorithm that leverages the repetitive nature of long tandem repeats.
• Analyzed k-mer frequency distributions within fixed-size windows to estimate potential repeat regions.
• Utilized a de Bruijn graph approach to assemble k-mers into a consensus repeat unit, overcoming high error rates.
• Compared performance against the established Tandem Repeats Finder algorithm.

Main Results:

• The proposed algorithm demonstrated superior sensitivity in detecting long tandem repeat expansions compared to Tandem Repeats Finder.
• The method effectively handles the high error rates inherent in long-read sequencing data.
• Successfully identified extensive repeat regions previously challenging to analyze.

Conclusions:

• The new algorithm provides an efficient and sensitive solution for detecting long tandem repeat expansions using long-read sequencing.
• This advancement facilitates the exploration of disease associations with large repetitive elements in individual human genomes.
• The open-source implementation (mTR) is available for broader research application.