search

MannLabs / alphapeptdeep

Deep learning framework for proteomics
Apache License 2.0
102 stars 20 forks source link

Adding mods into AlphaBase #120

Closed ehwmatt closed 7 months ago

ehwmatt commented 8 months ago

Hi,

I would like to train a model to predict RT/IM for peptides with a custom modification from DDA data (~10,000 modified peptide IDs output from MSFragger into a tsv library). I'm using PeptDeep via the GUI and am able to add my modification into "User-defined modifications" (and have it successfully update in the yaml file). I click to add the mod into AlphaBase but when I go back to the "Transfer" section to set up the training, the modification is not found in the AlphaBase mod search bar.

Please let me know if I'm misunderstanding the process to train a custom model for DIA library prediction of custom modifications and thanks in advance for any help!

Matt

jalew188 commented 8 months ago

Hi @ehwmatt , User-defined modifications will be automatically added into mod list, you don't need "the AlphaBase mod search bar" to add them again.

jalew188 commented 8 months ago

@ehwmatt Sorry, I have to double check this

Hi @ehwmatt , User-defined modifications will be automatically added into mod list, you don't need "the AlphaBase mod search bar" to add them again.

jalew188 commented 7 months ago

Hi @ehwmatt , for transfer learning, it is different as other search engines have different names for the PTMs you are interested. So you need to:

  1. Define your PTMs in User-defined modifications in AlphaBase format.
  2. Define the modification mapping of PTMs in other search engines to user-defined PTMs in Other modification mapping for PSM readers.
jalew188 commented 6 months ago

Closed by #125

ehwmatt commented 6 months ago

Hi @jalew188,

Thank you for your help with this!! Updating to 1.1.4 solved the first issue I was reported (where any mods I specified weren't shown as custom mods in AlphaBase for the mod mapping).

I still haven't got the GUI working on transfer/library mode but most likely user error I fear! Either way, executing an adapted version of the yaml from #129 from the command line works for trained custom models on my DDA data now and I'm in the process of evaluating the predicted libraries.

Thanks again! Matt