Multithreading blasting of unaligned reads

humann v3.9

Running data with humann3 with many unaligned reads increases the time of execution significantly. Recently I have used cProfiler to show that this time is primarily spent in the unaligned blastx portion of the software. If this section was multithreaded, would this decrease execution time?

Hello,

Sorry if I did not explain well enough. I had a cohort of deeply sequenced samples. Some of these samples ran for long periods of time, days, and in an effort to understand why these samples ran for so long I profiled one using python’s cProfiler (call graph above). It is fairly obvious. that according to the call graph, the majority of run time was spent in the unaligned blast portion of the humann3 application. I appeared to me that this portion could benefit from multi-threading, and I was looking at implementing this feature and wanted to reach out and see if this is something that would be accepted into the humann3 application.