Hi everyone 😊 we are currently developing several modules focused on metabolomics QC using MetaProViz and are planning to publish them to nf-core. To test these modules, we need to add new test data to the modules branch of nf-core/test-datasets. Since no metabolomics/ folder exists yet under data/, we wanted to check what the right structure would be (for example data/metabolomics/metaproviz/intracell_raw_se.rds?). Would that structure be the right approach, or is there a preferred convention we should follow? Thanks so much!

This a little tricky to say, from the point of view as it sort of depends on the field. E.g. genomics we had enough people agree to the structure of

<organism (i.e. genome>/<file format>

as that is how much of us generally work.

@Blanca Candelas Galavis?

Hi! Thanks for thinking of me. I'm mainly focused on metagenomics at the moment, so I'm probably not the best person to answer this one

Thanks so much for explaining that, it's really helpful to understand how genomics approaches this! Organism/format makes a lot of sense there. Our modules start from processed data, comparable to a feature count matrix in transcriptomics, meaning a data matrix (features x samples) and a sample metadata file. One addition on our end is a feature metadata file, since we often need several ID systems at once (e.g. HMDB, KEGG) to get enough matches for our features. In our case, all three are bundled into one SummarizedExperiment object saved as a single .rds file. Given that, we think a new metabolomics folder at the top level makes sense, but we were wondering if there is maybe an existing convention for processed data, perhaps already set up for count matrices, that we should follow instead? Thanks again for taking the time to help us think this through!

Yeah I agree, a metabolomics folder at the top does make sense. It's hard for my to say otherwise, as I have 0 experience of metabolomics... in genomics the processed data was always under the organism/format system (where within the format directory, you would have files derived from each other next to each other, e.g.

sample.bam

next to

sample.bai

where the

.bai

is an index file derived from

sample.bam

). I don't know what other structure we can have for findability purposes without more input from others. That said if you think your

.rds

file would only wokr specifically with MetaProViz and not other things, you can just use the exsiting

data/delete_me/<module_name>

structure instead of a dedicated metabolomics folder

Thanks for the input! We've decided on a

metabolomics/

folder, with our data in a

RawAnnotationData/

subfolder. Inside, we'll provide three TSV/CSV files (

DataMatrix

,

FeatureMatrix

,

SampleMatrix

), which is a standard format in metabolomics, not something specific to our modules. We'd originally considered bundling all three into a SummarizedExperiment

.rds

, but that's R-specific, so plain TSV/CSV keeps things usable regardless of language or tool. That's also why we think a separate folder makes more sense than

delete_me

, since it isn't tool-specific and is reusable for other use cases. On naming future additions: if someone wants to create a module that, for example, takes raw peak data as input (like the metaboigniter pipeline), we'd suggest creating a new folder named after what the data is, e.g.

metabolomics/RawPeakData/

, rather than after the tool it's used as input for, in keeping with your point about organizing by data type rather than tool/software. If anyone has concerns or suggestions for improvement, please let us know!

OK that sounds like a good plan, if RawAnnotation / TSV/CSV is standard 🙂 Please try and keep to the naming convention of snake case rather than Camel case though!

Thank you for the reminder! 😊