search

dcjones / proseg

Probabilistic cell segmentation for in situ spatial transcriptomics
Other
45 stars 3 forks source link

Overflow Problem in CosMx 6000plex #23

Open SanoKyohei opened 4 months ago

SanoKyohei commented 4 months ago

First, thank you very much for developing such a great tool!

Now we encountered overflow problem in CosMx 6000plex. This is error log.

proseg --cosmx Proseg_tx_file.csv.gz -t 10 Using 10 threads Read 644060072 transcripts 760384 cells 6718 genes Estimated full area: 126778810 Full volume: 1267788000 Using grid size 176.02158. Chunks: 7665 00:00:00 ############################################################ | log-likelihood: -5218807000 | assigned: 624538431 / 644060072 (96.97%) | non-background: (89.58%)thread 'main' panicked at /root/.cargo/registry/src/index.crates.io-6f17d22bba15001f/arrow2-0.18.0/src/array/physical_binary.rs:208:41: called Result::unwrap() on an Err value: Overflow note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

dcjones commented 4 months ago

Thanks for trying proseg. Just to verify, because I fixed a similar bug a while ago, is this with the latest version (1.0.5)?

Also, I haven't yet extensively tested proseg on higher plex. I think it may run into some problems in some cases, but I'm working on improving the model for these cases.

SanoKyohei commented 4 months ago

Thank you for your response. I was using version 1.0.3, so I will try with 1.0.5. By the way, are the differences between 1.0.3 and 1.0.5 only related to bug fixes and changes in output format? Is there a possibility of creating a batch effect by analyzing data output from both versions 1.0.3 and 1.0.5 together?

dcjones commented 4 months ago

No guarantees, but I think this bug was fixed with 1.0.5. There have been no changes to the actual model between 1.0.3 and 1.0.5, so it should not introduce batch effects.