Parallelize `umi_tools extract`

Hi,

I noticed that running umi_tools extract is taking a significant amount of time (~8 hours) with respect to the overall analysis pipeline that I have (200-300 million reads per sample/FASTQ). Could it be faster to first split the FASTQ, then run umi_tools extract with multiple threads and finally merge the parts together? I could probably do this manually in snakemake but I thought it would be more elegant to have it integrated into umi-tools, no?

Thanks, Roman

[vortexing]

I have umitools extract running and only needed the R2 reads. I ended up using a fifo to pipe the R2 reads from stdout from umitools directly to STAR to align those reads and the entire process is now taking half the time as before. So before I was saving R1 reads as a fastq.gz, and R2 reads as fastq.gz. So the lack of writing out the R1 reads, and the lack of requirement to compress either set of data is making this run MUCH faster. Just an FYI for future googlers. 😉

[paulranum11]

Just curious to know if any improvement to the umi_tools extract speed has been made or if any alternative tools are available that are faster. Any news?