CAREx: context-aware read extension of paired-end sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing (NGS) commonly produces short reads, limiting downstream applications.
• Longer reads are generally beneficial for various genomic analyses.

Purpose of the Study:

• To introduce CAREx, an algorithm for generating pseudo-long reads from paired-end Illumina short-read data.
• To enhance de novo assembly and other downstream applications by creating longer, more accurate reads.

Main Methods:

• CAREx utilizes repeated multiple-sequence alignments (MSAs) to extend short reads until their paired partners are identified.
• The algorithm is implemented in C++ (CPU) and CUDA/C++ (GPU), with a GPU-accelerated version available.

Main Results:

• CAREx significantly increases the number of connected read pairs compared to previous methods, especially on simulated data.
• It produces more error-free pseudo-long reads and achieves superior de novo assembly results.
• The GPU-accelerated CAREx demonstrates the fastest execution times among tested tools.

Conclusions:

• CAREx is an effective MSA-based algorithm for producing high-quality pseudo-long reads from short-read data.
• It surpasses state-of-the-art programs in read pair connection, error reduction, runtime, and de novo assembly.
• CAREx is open-source software available under GPLv3.