New pipeline: nf-core/eista

[huihai828]

Pipeline title/name

eista

Keywords

spatial transcriptomics, Vizgen MERFISH, 10x Xenium

What is it about?

EISTA is a bioinformatics pipeline that performs analysis for single-cell spatial transcriptomics data. The pipeline was developed as a generalized, flexible, and scalable workflow for spatial transcriptomics analysis. It is primarily designed for Vizgen MERFISH data and also support 10x Xenium data. The pipeline consists of three analysis phases. The primary phase includes Vizgen post-processing and converting count matrix into Anndata. The secondary phase focuses on single-cell QC, cell filtering, clustering analysis, and spatial statistical analysis. The tertiary phase involves downstream analyses, such as cell type annotation, differential expression analysis, cellular interaction analysis. Users can run the pipeline end-to-end or execute each analysis phase independently.

Please provide a schematic diagram of the proposed pipeline

I confirm my proposed pipeline will follow nf-core guidelines. Most importantly, my pipeline will:

• be built with Nextflow.
• pass nf-core lint tests and use standardized parameters.
• be community-owned and developed within the nf-core organization.
• open source under the MIT license with proper credits and acknowledgments.
• have a descriptive, all lowercase, and without punctuation name.
• use the nf-core pipeline template and predominantly use official nf-core modules.
• focus on a specific data/analysis type with appropriate scope.
• have properly maintained documentation.
• be bundled using versioned Docker/Singularity containers.

Why do we need a new pipeline?

The EISTA pipeline significantly enhances the reproducibility, scalability, and accessibility of spatial transcriptomics data analysis for Vizgen MERFISH. By automating primary to tertiary analyses including preprocessing, spatial statistics, and cell type annotation, etc. It streamlines workflows that are often fragmented or require substantial manual intervention. The modular design empowers users to run or exclude specific analysis stages flexibly, making it adaptable to diverse research needs.

Who would be interested?

Researchers and bioinformaticians working with spatial transcriptomics data.

What has been done so far

The pipeline has been under active development for few months, and has reached version 2.0. It has been tested within our institute, where we have validated its performance across multiple datasets and refined its features based on real analysis needs.

URL to existing work (if applicable)

https://github.com/EarlhamInst/eista

Are there any similar existing nf-core pipelines?

No response

[nf-core-bot]

Approval status: 🕐 Pending

Required approvals: Either 2 core team members OR 1 core team member + 1 maintainer

Review Status	Team members
🕐 Pending (Core)	@ewels, @maxulysse, @mribeirodantas, @mashehu, @JoseEspinosa, @mirpedrol, @FriederikeHanssen, @kenibrewer, @jfy133, @edmundmiller, @sateeshperi, @FranBonath, @christopher-hakkaart, @nvnieuwk
🕐 Pending (Maintainer)	@pontus, @maxulysse, @lpantano, @pinin4fjords, @mashehu, @JoseEspinosa, @mirpedrol, @GallVp, @FloWuenne, @matthdsm, @FriederikeHanssen, @adamrtalbot, @mahesh-panchal, @jfy133, @atrigila, @prototaxites, @ramprasadn, @edmundmiller, @luisas, @SPPearce, @vagkaratzas, @sateeshperi, @jonasscheid, @famosab, @nictru, @itrujnara, @LouisLeNezet, @Joon-Klaps, @jasmezz, @Felix-Kummer, @sofstam, @nvnieuwk

[JoseEspinosa]

Does the proposed pipeline have any overlap with spatialvi? The feedback from @FloWuenne and @ewels might be helpful for this proposal 🙏

[FloWuenne]

Hi @huihai828 ,

thanks for your pipeline submission. So there are a few pipelines with some overlap in nf-core: nf-core/sopa (not released yet) and nf-core/molkart, nf-core/spatialvi and nf-core/spatialxe. From looking at your metro map, it looks like there is some overlap with SOPA and spatialXE (since you are processing Vizgen or Xenium data).

Could you please elaborate how your proposed pipeline distinguishes from the approaches implemented in these pipelines and which of the downstream tools overlap and which ones would be completely new? I think it's important to know how your Xenium processing here would be different from the spatialXE pipeline and if there isn't, why a separate pipeline is required. Can you also elaborate on the name for the pipeline? We generally encourage pipeline names in nf-core that reflect what the pipeline does or which assay it's used for (molkart = Molecular Cartography, Spatialxe = spatial xenium etc). What does eista stand for?

[huihai828]

Hi @FloWuenne, thank you for the comments!
The pipeline name EISTA stands for Earlham Institute Spatial Transcriptomics Analysis. It was originally developed for Vizgen data analysis used within the Institute. I have reviewed the SOPA and spatialXE pipelines which are also capable of processing Vizgen data for cell segmentation and count generation. The main difference is that EISTA uses the VPT toolset, configured with its segmentation algorithm file. The pipeline is currently able to read Xenium data through the spatialdata package. In addition, EISTA provides some downstream analyses. Given the overlap with existing pipelines, I agree that EISTA appears somewhat redundant.

[FloWuenne]

Thanks for clarifying the overlap @huihai828. The reason I am asking about overlap is that we try to keep pipelines in nf-core generally unique per assay / data type / analysis. This is obviously extremely difficult with spatial analysis pipelines, where there is a lot of overlap between methods like segmentation, processing and downstream analysis.

To me it looks like the only other pipeline that could potentially be able to analyze Vizgen data would be SOPA. Is the overlap with SOPA sufficiently big that you think EISTA would just replicate a lot of functionality or is there sufficient functionality within EISTA that is unique, that you think it differentiates from SOPA? I think the main thing we want to make sure of, is that it's clear what each pipeline is for. If your pipeline can be a more Vizgen specific pipeline and SOPA doesn't solve that same Vizgen processing need, I think that would be a convincing avenue to pursue. Let us know what you think! 😊

[huihai828]

Thank you @FloWuenne for the additional information, which is very helpful for me to better understand the philosophy of nf-core. I can see that SOPA is designed to be more general and capable of handling a wide range of image-based spatial transcriptomics data. In contrast, EISTA is designed specifically for Vizgen data, since it uses the Vizgen Post-processing Tool (https://github.com/Vizgen/vizgen-postprocessing), which is purpose-built for processing Vizgen MERSCOPE data, updating .vzg files, and configurating segmentation algorithm file which defining all algorithm-specific parameters (model type, thresholds, channels, z-layers, tile size, etc.). In addition, we provide some common downstream analyses, such as spatial statistical analysis with Squidpy, which revealing relationships between gene expression patterns and tissue morphologies. We also plan to extend the pipeline with more spatial-related analyses. These features could form a unique selling point for EISTA for Vizgen data.

[FloWuenne]

I agree that focusing on the Vizgen processing aspect of the pipeline as it's differentiator is a good idea. I would also suggest to rename the pipeline for nf-core to something that better reflects this spatial vizgen purpose. Some ideas, remember it will always be nf-core/ in front of it due to the org being nf-core, so read all of this as nf-core/name:

• spatialvizgen
• spatialmerscope
• merscope

[huihai828]

Great, thank you @FloWuenne! I like the name nf-core/spatialvizgen, and I’m happy to maintain spatialvizgen as a separate repository under github.com/EarlhamInst for nf-core. I have one question: would it be fine to keep my institute’s name in the README and the HTML report?
Once everything sorted out, could you please let me know how I transfer the spatialvizgen repository from EarlhamInst to nf-core?

[FloWuenne]

So these are all just recommendations from my side. What would be required to make your pipeline an actual nf-core pipeline is to get approval here for your pipeline being part of nf-core. The goal of nf-core is to provide standardised, best-practice pipelines. You can check the Guidelines here: https://nf-co.re/docs/guidelines/pipelines/overview.

Once you get approvals for your pipeline to be included, you would either transfer your existing repository (if you started it with the nf-core template) or we would generate a new repo under nf-core with the current template version. You would then implement your pipelein according to the best practice standards as defined by nf-core and work towards a first release.

Let me know if you have any additional questions.

[vagkaratzas]

This is a difficult one. With a quick look, it seems like nf-core/sopa is a superset of nf-core/spatialvi and nf-core/spatialxe mentioned so far. If vizgen data/modules are added to sopa, then there will also be overlap between that and the proposed spatialvizgen. So two options here;

1. create a new pipeline with possible future overlap
2. contribute directly to the existing sopa pipeline, updating the samplesheet and adding the missing subworkflows

I guess another reason to create a new pipeline would be if sopa only does high-level computations (broad but shallow) for all different data types, whereas the rest of the spatial pipelines go into more detail depth (narrow but deep).

[huihai828]

Thank you @vagkaratzas for the review! Yes, I agree that SOPA already covers a wide range of image-based technologies and allows setting CLI parameters for tools like cellpose, and may unnecessary to add other similar toolsets like Vizgen Post-processing Tool (VPT) since itself wrapping on tool cellpose. This helps avoid redundancy for nf-core. Probably I would propose an nf-core module for VPT for anyone to use. Thanks!

[FloWuenne]

@huihai828 To follow up on this, do I read your comment correctly that because of overlap with SOPA you think its better to just add an nf-core module for Vizgen and not implement this pipeline as nf-core? If I misunderstand your message, sorry, I just wanted to clarify!