Hybrid de novo tandem repeat detection using short and long reads

Area of Science:

• Genomics
• Bioinformatics

Background:

• Tandem repeats are significant genome rearrangements impacting inherited diseases.
• Reference-based tandem repeat detection is effective, but de novo detection remains challenging.
• Short reads lack the length to span long repeats, while long reads have high error rates.

Purpose of the Study:

• To develop a de novo method for tandem repeat detection that overcomes limitations of short and long reads.
• To address the challenge of high error rates in long-read sequencing data.

Main Methods:

• MixTaR is a hybrid algorithm combining short and long reads for de novo tandem repeat detection.
• It utilizes de Bruijn graphs with short reads for pattern detection.
• Long reads validate patterns, and local greedy assemblies construct repeat sequences.

Main Results:

• MixTaR was tested on simulated and real data from complex organisms.
• The method's robustness to varying error rates in short and long reads was analyzed.
• Results were evaluated based on the number and pattern lengths of detected tandem repeats.

Conclusions:

• MixTaR demonstrates high precision and sensitivity in tandem repeat detection.
• The method achieves low false positive rates, even with highly erroneous reads.
• Accurate detection of tandem repeats with variable pattern lengths is achieved.