Refactoring of lesSDRF

[TineClaeys]

Refactoring of lesSDRF: Feedback welcome!

I've pushed a refactored version of the lesSDRF app to the refactor-streamlit branch. Based on a lot of discussions, this refactoring aims to make high-throughput SDRF annotation faster and more intuitive.

Try it here: refactorlessdrf.streamlit.app

What’s new:

• Mapping local metadata and labels requires less clicking
• Annotation of columns should be more streamlined
• Allowing ontology based on term matching and using it to fill in the SDRF
• Community suggested columns for subniches such as single-cell proteomics or immunopeptidomics

I'd love some feedback! Please test it out and let me know:

• What works well?
• What’s confusing or missing?
• What would help your workflow?

Feel free to leave comments, issues, or suggestions directly here.

Validation is already being worked on ;)

[Cajac102]

Stuff I've ran into (somewhat chronologically ordered):

• Labeling:

1. I selected only label free, but two columns were created in the preview: "Assigned label 1" and "Assigned label 2"
2. When I selected 6 TMT labels, also only two columns were created, so I can only put in 2 labels per file
3. The "Select ALL" button is missing

What would help: If I could copy labels from one file to the other (e.g. TMT16plex -> i would have to select 16 labels for each file)

• Annotate columns:

1. The menu says "Annotate columns", the tab is called "required columns"
2. I love the progress bar
3. Maybe switch optional and required so that you can work your way from top to bottom?
4. Feels a bit clunky to have a list of all the required columns I still have to fill, but not to be able to click on it and instead having to use the drop down menu below
5. I have an "AG Grid - for trial use only" watermark sometimes in the table preview
6. Would be more easy if I was redirected directly to the next required column instead of having to select one each time (17 times with what I tried out)
7. In the assay name annotation, I have the text "Selected column type: <class 'str'>", which does not look like it should be there?
8. For some columns, the text tells me that I HAVE to input numerical values, but I can input text 😈
9. A lot of text in modification parameters. would be nice if it could be broken down step by step with a bit of formatting to ease reading
10. Fragment mass tolerance could have a dropdown menu for Da or ppm so that the formatting is not left to user (10 Da? 10Da? 10DA? 10 Dalton? 10.0 Da?)

-Ontology input:

1. Love it
2. Some guidance towards what is characteristics and what is comment when selecting for a new column would be nice
3. After I click update, the table on the top does not show the new input :( it works in the Experiment type integration though

Is the download available after validation? I can't find the button
Did not run into any hard errors (except for validation but that's the expected one)

Hope that helps 💌

[mobiusklein]

Tested in Chrome, on Windows 11:

Local metadata mapping

• While mapping local metadata, selecting the mapped column name would immediately erase the value entered when entered for that column for the first time. The second time I enter the name to map to works as expected, but this applied to every single column I mapped.
• Some of the columns in my local metadata file corresponded to columns not in the subset that were mappable and that information was not propagated to the next stage.
• I was not sure what format I was expected to upload this data in and how it had to relate to my starting data. After reading the source code, I could see it was going to just do a horizontal concatenation.

Annotation

• I observed the same issue with clunky repetitive re-selection as previously described, and with the odd type object being printed out: . The type printout might be debugging related?
  

  lesSDRF/pages/3_3. Annotate_columns.py

  Lines 318 to 321 in 1ff3ae7

  	if selected_column in FREE_TEXT_COLUMNS:
  	st.write(f"Selected column type: {type(selected_column)}")
  	st.write(f"{extracted_name} requires free text input annotation")
  	builder.configure_column(selected_column, editable=True,cellEditor="agTextCellEditor",cellStyle={"background-color": "#ffa478"})

• For large studies, there was no way to set a single column to a value, like for when a value isn't available, but the column accepts free or nearly free text. AgGrid lets you drag-fill, but the browser really didn't like scrolling with that interaction enabled. A work-around for this was to hide all but the column I wanted to fill, enter the value in the first row, begin drag-filling, and then press page-down to jump to the end of the table (quicker since there is less for the JS-heavy table widget to tell the browser to render), and then finish the drag.
• Error when trying to enter dissociation method:
  
• Programming error when trying to validate the SDRF file: