IsoQuant WorkTime

[laetitiarialland]

Hello,
I was wondering how long it takes to run a ONT-long read for a single sample (from bam file with 8GB size) ? I've been trying to run this sample and it appears to work (no error) but it is very long (running for 200hrs in 128gb RAM ; and it is not finished (step : Processing chromosomes)). I wonder if it is normal ; or if you are aware of some bugs when you start from bam file or with only one sample ?

Thank you,
Kind regards,

Laëtitia

[andrewprzh]

Dear @laetitiarialland

It's hard to say in general. For example, I just found a log in which a 37 GB BAM file was processed in 3 hours in 16 threads. However, I encountered cases in which a lot of reads map to the same region (sometimes mitochondrial chromosome) which causes slow processing. I don't think there are any particular issues with running a single sample or using a BAM file. By the way, which aligner did you use?

Performance was significantly improved in the recent version, so make sure you are using the latest one. Disc I/O can also be an issue. Could you also check how many processes are actually running?

200 hours seems like quite a lot, could you send me the log file as well?

Best
Andrey

[laetitiarialland]

Hello, thanks for your answer.
It could be linked with mitochondrial chromosome because I've a lot of reads that are mapping to this region. I use minimap2 for alignment.
The version I have is 3.2.0, I will try to upload to 3.3 to see if it works better.
Here is the log file for the run :
isoquant.log

Thanks
Kind regards
Laëtitia

[andrewprzh]

Dear @laetitiarialland

Seems like chrM is not the issue in this case. There are several chromosomes that being processed for too long. As I mentioned, the problem is typical for genes/regions with extreme amounts of reads. If you have a chance to send me a part of your BAM file, let's say chr9, I can take a look at it as some point.

And unfortunately 3.3 should not be different in terms of performance.

Best
Andrey

[dawangran]

Hello, I have 17 samples, and I can finish two samples in three days, which consumes a lot of time, is there any way to solve it?
isoquant.txt

[andrewprzh]

@dawangran

Running time is not always proportional to the number of sample or the number of reads. The most problematic cases are loci with extremely high coverage.
For example, in your case it is the mitochondrial chromosome that requires 3 days alone (the rest are processed within hours).

2024-02-12 19:33:57,258 - INFO - Finished processing chromosome 4
2024-02-15 09:27:41,671 - INFO - Finished processing chromosome MT

It is a known problem, which I working on, but it is not easy to deal with without changing the functionality.
The only workarounds now are either to remove MT chromosome if you don't really need it, or use --no_model_construction option, which will disable novel transcript discovery.

By the way, if you could share a BAM containing reads mapped to MT, that can also help.

Best
Andrey

[andrewprzh]

Finally released new version 3.4, which has significantly better performance, especially for such cases.
On my test dataset new release version works for 3-4 hours instead of 3 days.

Best
Andrey