CLC bio's read mapping algorithm beats popular open source alternatives in benchmarks

VIENNA--(BUSINESS WIRE)--Today at the Bio-IT World conference in Vienna, CLC bio officially released the white paper on its new read mapping algorithm, which is included in the current releases of CLC Genomics Workbench version 5.5 and CLC Genomics Server version 4.5. The white paper covers four human datasets from the Illumina GAII, Roche 454 Titanium, Life Technologies Ion Torrent PGM, and PacBioRS platforms, and benchmarks CLC bio's read mapper for performance and accuracy against the open source read mappers BWA, Bowtie2, and SMALT.

Highlights from the white paper include mapping of almost 99% of 1.34 billion paired-end Illumina reads in 5 hours, compared to 28 hours and 99% for Bowtie2, 41 hours and 96% for SMALT, and 60 hours and 95% for BWA. CLC bio's new read mapper scales extremely well with multi-processor machines, enabling up to 35-fold decrease in time to finish a read mapping, and delivers the highest accuracy in almost all the benchmarks.

Program Director of World-wide Deep Computing in Life Sciences at IBM, Janis E. Landry-Lane, states, "IBM has been productively engaged with CLC bio for nearly two years. This collaboration has focused on optimizing the performance of the CLC Assembly Cell de novo and reference mapping. We have achieved remarkable results, and, I believe, performance of this order will be required for the rigorous demands of genomics research. We look forward to serve our customers together."

Bioinformatics Specialist at CLC bio, J.-Uwe Appelt, PhD, adds "PacBio is popular for de novo projects, as its long reads are ideal for scaffolding and currently BWA is considered the best open source read mapper for PacBio data. However, this benchmark white paper proves that our new mapper, mapping 46% of the reads correctly, outperforms BWA which got 5% correct and resulted in a high false discovery rate. One of the reasons behind the strong performance of our new read mapper is that it automatically adapts to arbitrary error-modes, which means the users don't even have to tweak the parameter settings to get the best results."