How do alignment programs perform on sequencing data with varying qualities and from repetitive regions?

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing (NGS) generates numerous short reads for biological research.
• Sequencing reads often exhibit low quality at the 3' end and originate from repetitive genomic regions.
• Performance variations of alignment algorithms under these conditions are not well understood.

Purpose of the Study:

• To systematically evaluate the performance of four common alignment algorithms: SOAP2, Bowtie, BWA, and Novoalign.
• To investigate the impact of read quality and repetitive regions on alignment accuracy.
• To compare algorithm concordance using real and simulated sequencing data.

Main Methods:

• Utilized real and simulated next-generation sequencing data with varying quality and from repetitive regions.
• Assessed alignment algorithm performance based on concordance between aligners (real data) and alignment accuracy (simulated data).
• Evaluated the effect of trimming low-quality bases on alignment outcomes.

Main Results:

• All four alignment programs (SOAP2, Bowtie, BWA, Novoalign) performed similarly on high-quality or trimmed data.
• Trimming low-quality ends significantly increased aligned read counts and inter-aligner consistency, particularly for low-quality data.
• Novoalign showed increased sensitivity to data quality improvements; trimming enhanced its concordance with other aligners.
• Simulated data indicated that reads from repetitive regions are prone to incorrect alignment, and filtering multi-hit reads improves accuracy.

Conclusions:

• This study offers a comprehensive comparison of popular alignment algorithms for sequencing data with quality variations and from repetitive regions.
• The evaluation methodology is adaptable for diverse sequencing datasets and platforms.
• The approach can be extended to assess the performance of other alignment programs.