Another PacBio Development - Adam Phillippy's New MHAP Module

Alignment speed is the biggest bottleneck in PacBio assembly. Therefore, those working on PacBio reads will find the following release helpful.

June 14, 2014 - PBcR and CA 8.2 alpha as source or pre-compiled for Linux is now available. PBcR now incorporates a novel probabilistic overlapper for self-correction of sequences named MHAP. This allows assembly of prokaryotic genomes in < 30 minutes on a typical desktop and assembly of small eukaryotic genomes in < 2 days. If you use MHAP, please cite the Biology of Genomes poster (Berlin K., Koren, S. et. al. Reducing assembly complexity of genomes with single-molecule sequencing. Biology of Genomes, 2014). For best results, java 1.7r51 or newer is recommended to use MHAP.

How good is it? Here is the plain language description -

-———————————————

Given that alignment is a big bottleneck, we previously checked whether BWA- mem could improve the execution time. Heng Li came up with a set of optimal parameters to get the best alignment. Readers may find the following comment from Irek in that thread helpful -

Hey, I just finished comparison for new PacBio RSII data (CCS and CLR), used: bwa-sw,mem,blasr,ssaha2,smalt,last and agile.

Checked speed, memory, mapping status of reads, then went for precision-recall assessment, and finished with the analysis of error model recognition.

Actually new version of SMALT looks like a winner and its really fast and memory efficient.

As for mem blasr. For CCS they are comparable in terms of precision-recall, but for CLR, mem definitely looses.