HDExaminer conversion and allow replicates

[tuttlelm]

models.py
Changes were made to the HDXMeasurement init function to allow the original HDX-MS data to contain replicates.
HDExaminer output can contain either summary data for every fit spectrum or replicate averaged data. This change allows creation of the HDXMeasurement object from either type of data.

convert_data.py
Function to convert the summary uptake table output from HDExaminer to a format usable by pyhdx.
This primarily involves mapping column names, but will sometimes require dummy values (e.g. if rfu_sd is missing, a 5% error value is assumed)

[Jhsmit]

Thanks for the PR!

Unfortunately it look like some of changes to models.py cause issues for the web application part of PyHDX .
I'll try to take a look but I'm not sure if I'll have time soon due to holidays.

If you have an example 'all_results.csv' file that would be very helpful so we can add automated tests.
It doesnt have to be real data, just as long as the format is correct we can use it for testing.

[tuttlelm]

Unfortunate that it breaks things in the web interface. I have only used pyHDX functions in a Jupyter Notebook. I've added a couple example HDExaminer outputs in HDExaminer_examples. You can see that the HDExaminer outputs aren't even self-consistent in column labels. Both formats are handled by the convert_data.py function.

[Jhsmit]

Hi, I've added support for HD examiner now in the latests release
PyHDX should now support any openHDX compatible format as defined by hdxms-datasets

However, all input gets aggregated across replicates. I think your original use case involved calculating residue-level weighted averages for replicates individually?

If so, I might release something in the future which can do that.

[tuttlelm]

That's great, I'll try to test it out with my workflow when I get a chance.

I do keep replicates separate in my analysis. There's maybe two main reasons for this. One is just general quality control - your coverage plots are a great way to qualtitatively check for outliers, and if you create the hdxm measurement object on the already averaged data then you lose some of that. The specific motivation to keep replicates separate in my multimodal anlalysis is that sometimes a replicate will better fit a unimodal distribution than bimodal or vice versa and I want to keep that information.