[Feature] Bypass Baysor patches where no cells

[lguerard]

Is your feature request related to a problem? Please describe.
When running sopa using snakemake, everything runs fine until some Baysor patches fail due to not enough cells pass prior min_mol_per_cell. This can definitely happen if some patches are outside the sample or low transcripts in specific regions. However, this then blocks the pipeline from continuing.

Describe the solution you'd like
The API has the possibility to get the valid_indices for the patches to continue, is there any way of using that for snakemake as well ?

Describe alternatives you've considered
If this is not possible, would it be possible to move out/rename these patches and continue with the pipeline ?

Thanks a lot for your help !

[quentinblampey]

Hi @lguerard,

Actually, the snakemake pipeline is already using the valid_indices. But it's difficult to check whether Baysor will succeed or not, and currently we decide to keep a patch if there are more than 4000 transcripts in that patch.

In sopa==2.0.0, there will be an API argument to continue running baysor even when this happen (force argument), I'll let you know when it's released!

[lguerard]

Thanks for the (very !) fast answer !

So the slurm job exits with an error which then stops the whole pipeline, is that expected ? It also doesn't resolve if you rerunning it as it's looking for the loom files (makes sense as it doesn't have the valid_indices anymore.

[quentinblampey]

Normally, 4000 transcripts should be enough for Baysor, but sometimes, there can still be an error. Indeed, in that case, the pipeline stops, and you'll not be able to finish running it except if you remove the index of the patch from the valid_indices file from the cache (which is quite painful to do).

So hopefully, the force argument from the API will help, I'll let you know :)

[lguerard]

I see, I'll try and investigate why it failed this time and move/change things manually while waiting for 2.0. Looking forward to it, thanks a lot for your awesome work! Le mar. 5 nov. 2024, 09:20, Quentin Blampey ***@***.***> a écrit :

…

Normally, 4000 transcripts should be enough for Baysor, but sometimes, there can still be an error. Indeed, in that case, the pipeline stops, and you'll not be able to finish running it except if you remove the index of the patch from the valid_indices file from the cache (which is quite painful to do). So hopefully, the force argument from the API will help, I'll let you know :) — Reply to this email directly, view it on GitHub <#150 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABPFJWZ2GIBJIEM26VY7IMLZ7B5UJAVCNFSM6AAAAABRESWXXGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJWGUYTKMZQGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>