W-L
commented
1 year ago Hi Ben,
Thanks for your efforts and for the detailed logs, that's very helpful. From a first look it seems that it should be working fine. As you say, BOSS-RUNS creates new masks and readfish picks them up as well. At least at the very beginning of the sequencing run. Do you see any additional Reloaded mask dict for.. in the readfish log? That would indicate that the masks from BOSS-RUNS very successfully replaced by updated ones. I'm asking since it seems that the masks that BOSS-RUNS produces instruct quite a few reads to be rejected. E.g. from what I can see, reads mapping to the reference CLASSIFIED_HEADER_NAME_25.1 make up about 47% of your sample, and at the time of the second log you posted as EDIT, ~55% of the sites are classed as "to be accepted" if a read starts there. So if this single sequence makes up ~50% of the sample and we reject ~50% of it, we should see at least ~25% of total reads getting rejected. In readfish's chunk.log these should appear mostly as single_off unblock instructions, as you probably expected. The part of the chunk.log you posted seems to be the very top of the file from the start of the experiment. Did you check if any single_off unblock select_species were recorded later on? You can grep for them with grep -P "single_off\tunblock\tselect_species" chunk.log, since they would appear just like any other read unblocked by readfish. If using BOSS-RUNS, readfish has no other masks that it works off of, so there is no special tag or anything to distinguish which reads were rejected because of BOSS-RUNS.
Besides that, what is the distribution of read lengths in your library? Did you see a peak of reads being rejected? Or any difference between the reads from the control and adaptive sampling sectors on the flowcell?
Thanks!
bnoordijk
Hi!
We've used Readfish in the past to eject unwanted reads; with great success! Now we want to start using BOSS-RUNS to prevent a coverage bias in our sample. We've configured it according to documentation, and ran it successfully on the simulated human read set.
However, when we run it on a real sequencing sample, we're not sure if BOSS-RUNS ever tells readfish to eject any reads. We compare it to one control zone, and sequence on a MinION flowcell.
Our toml looks like this:
We run readfish and bossruns with the following commands:
and
Each command is run from within a docker container, but the explicit command which mounts all the relevant volumes is not shown here.
It seems like boss-runs can update the masks dynamically; see this excerpt from our BOSS-RUNS log:
Readfish also successfully reloads the mask dict:
Excerpt from the readfish chunklog. Readfish only eject reads because they exceed max_chunks it seems.
How can we find out when Readfish ejects reads because of something to do with BOSS-RUNS?
The sample we sequence consists mainly of one species. We use readfish because we want more even coverage across our genome. But the results of our experiment do not indicate any difference in coverage. Do you have any ideas what could cause our issue?
We're happy to provide more information if you need any.
Kind regards, Ben
--EDIT-- Do you maybe think it has something to do with BOSS-RUNS detecting many dropouts? An example can be seen here: