goranrakocevic
goranrakocevic
Forgot to say: - For making the graph we just used autoindex with the fasta + vcf - For mapping we are using giraffe with `-b fast` (or without, same...
A small update, if it helps: if I reduce `--hard-hit-cap` to 100 the problem seems to go away (i.e. mapping is very fast). I tried a few other parameters, this...
Yes, the fastqs are public (1KG data), this sample: https://www.ebi.ac.uk/ena/browser/view/ERR3239276?show=reads It doesn't seem to be a single read, because writes to the output gam slow down (didn't check the code,...
They do map, just take a long time. Is the time saved by default? (Ie we can do it on any gam)
Hi All, Sorry for the long silence. So far it seems we don't see very few reads that are very slow to map, but a larger number of reads somewhat...
While at it, I think I see another small issue. - mmseqs_search_monomer calls unpackdb with --unpack-name-mode 0 (thus creating 0.a3m, 1.a3m, etc. for each chain) - Later on the script...
If you look at the logs, you'll see that the slowest steps are not the ones using the GPU, but steps that come after, that are very IO (disk) intensive....