Tool:Sambamba: High Performance Alternative For Samtools And Picard For Indexing, Sorting And Merging Bam Files

Question

[William]

Sambamba is a high performance modern robust and fast tool (and library), written in the D programming language, for working with BAM files. Current functionality is an important subset of samtools functionality. Because of efficient use of modern multicore CPUs, usually Sambamba is much faster than samtools. For example, indexing an 18 Gb BAM file on a fast 8 core machine utilizes all cores at 45% CPU:

Sambamba index bam:

time ~/sambamba index /scratch/HG00119.mapped.ILLUMINA.bwa.GBR.exome.20111114.bam            
real    1m42.930s
user    6m19.964s
sys     0m32.362s

Samtools index bam:

time ~/samtools index /scratch/HG00119.mapped.ILLUMINA.bwa.GBR.exome.20111114.bam 
real    5m37.669s
user    5m9.127s
sys     0m13.605s

https://github.com/lomereiter/sambamba

[Aaronquinlan]

How many threads were used for the sambamba time?

[Matt Shirley]

I would also like to know how many concurrent threads were used, but assuming only the userspace code was multithreaded we can do (real - sys) / user which is approximately 5. If the ~45% utilization figure is correct, then 5 * 1.55 = 7.75, so approximately 8 threads.

[William]

Completed the quote with the thread info.

[Christian]

Isn't disk IO the main bottleneck in this operation?

[William]

I guess that depends on the storage setup used. The faster the storage you use, the more the speedup is (see results for indexing). https://github.com/lomereiter/sambamba/wiki/Comparison-with-samtools

[elvissober]

How to install and use it correctly? I have:

user@user-MS-7817:~/Documents/sambamba/sambamba$ make sambamba-ldmd2-64
make: *** No rule to make target `sambamba-ldmd2-64'.  Stop.
user@user-MS-7817:~/Documents/sambamba/sambamba$