The possibility of MPI implementation of StrobeAlign

[jianshu93]

Hello Kristoffer,

I saw an interesting reimplementation of BWA-MEM Using MPI (Message Passing Interface) here: https://github.com/bioinfo-pf-curie/mpiBWA

It seems a multi-node solution is becoming more appealing these days because most research groups have access to institution-hold HPC systems with thousands of computing nodes. The master process in MPI can just split reads into pieces and assign them to other processes on each node and collect (wait) results after all are done. I tested the mpiBWA with 10 node or something, a 25G Metagenome can be mapped to 500M genome in several minutes with exactly the same results with single node bwa mem. I think a mpiStrobeAlign will be insanely fast using 10 or so computer nodes.

Thanks,

Jianshu