protdata to load from common search engines

[mffrank]

Hi openDVP developers,

I work with subcellular proteomics at the Chan Zuckerberg Biohub in San Francisco and I'm using anndata and the scverse ecosystem for my analysis. I wanted to have a generic way to load the output of different search engines into AnnData, so I created protdata (docs). It's a very small package with minimal dependencies that just does IO.

I recently met Fabian at ASBMB and saw that you're also working with AnnData and spatialData. Do you think it would be useful to collaborate on protdata and make openDVP use it for IO? As far as I can tell it should pretty much be a drop in replacement. Happy to chat and submit a PR if you think it would be useful!

Best,
Max

[josenimo]

Hey Max,

Thank you for reaching out!
protdata looks really great!

I think there is potential for such an io interface especially with all the various formats.
There are some features in DIANN_to_adata.py that are not present, such as:

1. Removing protein groups labelled as contaminants
2. Shortening Gene lists to first gene (a poor but necessary step for things like pathway enrichment).
3. Parsing of a metadata file (report.stats.tsv a common one around here), or custom metadata file with sample info. This could honestly be another function to add metadata to the adata object to keep functionalities separate.

What are your thoughts on these? we perhaps generalize this for the various input formats?

Unrelated but important, I was talking to HUPO-PSI about making AnnData the default Quantification matrix for LCMS-based proteomics, since there is currently no standard. Would you like to participate in the discussion? I think it would be great to add more opinions :)

[mffrank]

Hey Jose,

Thanks for getting back so quickly!

I agree that preprocessing is missing so far. I wanted to keep the scope super minimal initially, but you make a good point that this should probably be in protdata since it's dependent on the input format (e.g. some search engines might have different separators for gene lists).

I have some of this functionality in my main package grassp for now. I think it makes sense to port them over! I'll start working on the contaminants and gene lists and qc metadata.

And would love to join for the HUPO discussion. I also have a colleague at the Biohub, Carlos Gonzales that is pushing for something like this and would be great to include. I'll DM you in the scverse zulip chat.