Their efforts align very well with spatialxe

Baysor is now available as a bioconda package, I will make other packages part of bioconda, so in the next verison, we might be able to support conda for spatialxe.

We waited 2 years just to get it into conda with the baysor binaries 🫠😵‍💫🙂‍↕️ in the end

@Felix Krueger / others - are you still keen on getting this done? https://github.com/MultiQC/MultiQC/issues/3348 Just taking a look at it ahead of the v1.35 release. Wondering a bit if it should go into MultiQC or into the xenium-extra plugin.

I think we need to come up with a way to really reduce the

GPU vram

usage for cellpose. We have logically partitioned gpus (24gb each) and the cellpose jobs failed on those. I had to request an entire

72gb nvidia_h200_mig_3g.72gb

gpu just to run 1 cellpose job and that consumed around 68.5Gb vram.

I looked at the cellpose cli options if any and the

batch_size

is already implemented. The other is

resampling

which reduces segmentation quality

Not sure if you guys have the same problems

most tools/packages based of of pytorch do not use so much vram. Cellpose somehow stands as an exception

let me know if a region based segmentation is possible in cellpose and if this can be implemented in spatialxe

this is how vram usage looks as cellpose is running

mmm that would have been my first guesses to optimize.

Maybe this needs to be implemented:

cellpose.contrib.distributed_cellpose

I had a discussion with 10x and atera will have slightly differetly input format than xenium. but only slightly, because of data size. So depending how quickly SpatialData catches up we could decide to either wait for a SpatialData update or implement directly a support for the atera data type bundle.

<!channel> We have started discussions on how to handle the 10x Atera data. At the moment, we are considering two possible approaches: 1. Creating a new dedicated pipeline, or 2. Extending the existing spatialxe pipeline with merged functionality, potentially under a new name. To help us make an informed decision, I would like to organize a meeting in the upcoming days. It would be great to have 1–2 people from nf-core join the discussion as well. This is also relevant for the upcoming release for spatialxe. Please indicate your availability here (Times are chosen to covere US (PDT) and Europe (CEST)): https://terminplaner6.dfn.de/p/af34e62b7d17c357bd7e2887f15d1403-1761456

Dear all we will rename the pipeline to account for the fact that it will support both 10x Xenium and Atera data. For that we already had ideas: 1️⃣ spatialax 2️⃣ spatialatxe 3️⃣ spatialaxe Which one would you like better? We are also open for other suggestions.

has renamed the channel from ‘spatialxe’ to ‘spatialaxe’

If anyone is using baysor in the conda profile, please remove the cached conda env as there was a new build for baysor pushed to bioconda today

the version stays but the recipe was updated with a minor fix

Hi @Florian Heyl @Dongze He,

proseg

,

baysor

and

opt

are now available as

nf-core modules

and can be installed in spatialaxe. These also run with

bioconda

now. We can add these in the next version of the pipeline.

Opt subworklfow

is also part of the nf-core subworkflows that can be installed too. This would also enable us to add the conda profile back in the pipeline.

wonderful! Thank you very much for working on it! I will find some cycles to update these modules!

<!channel> the first release PR was merge with PR 139 🥳 It still needs some little things before it is official official.

With that, the release is official 🎉. For those who were directly involved, thank you so much for the hard work 💋! For those who were supporting with reviews, suggestions, and PRs that did not make into spatialaxe yet, we appreciate your hard work 💋! Thank you so much! We will continue of course developing the pipeline. If you find bugs or any problems, please open issues. The spatial omics field is still moving fast, and it is sometimes hard to keep up with the new changes.

Thank you @Florian Heyl @Dongze He for the efforts. @Kübra Narcı @Maxime Garcia @Phil Ewels @mashehu for all the PR approvals, container problems and other nf-core stuff.🚀

Congratulations all! Huge milestone! 🎉

message has been deleted

We tried to run this pipeline on a publicly available Xenium dataset using Seqera Tower and AWS batch. It seems to run into the following file check error when running on AWS batch. Is this expected behaviour?

Missing required file(s) in xenium bundle '<s3://xxxx/data/spatial_transcriptomics/xenium/public/mouse/GSE272303-SNCA-tg/Line61nTg_1/outs>': [cell_boundaries.csv.gz, cell_boundaries.parquet, cell_feature_matrix.h5, cell_feature_matrix.zarr.zip, cells.csv.gz, cells.parquet, cells.zarr.zip, experiment.xenium, gene_panel.json, metrics_summary.csv, morphology.ome.tif, morphology_focus/, nucleus_boundaries.csv.gz, nucleus_boundaries.parquet, transcripts.parquet, transcripts.zarr.zip]

Hey! Spatialaxe is failing on the full size tests: cloud.seqera.io/orgs/…/logs Could someone take a look please? It would be great to get example results up on the website

For Hackathon, modules to tackle would be: SPLIT, grandqc, DenoIST, troutpy.

SeqTraQ can also be tackled but there is already a module, yet it needs runtime optimization.

image.png