andrewrech / antigen.garnish

Other
45 stars 13 forks source link

garnish_antigens filtering #144

Closed nickhir closed 2 years ago

nickhir commented 2 years ago

I have a few questions regarding the garnish_antigens filtering. The documentation says, that the tool returns a data frame with ranked neoepitopes. I was wondering if you could elaborate how exactly the ranking is done, e.g. if certain criteria's are weighted more than others. Furthermore, I was wondering why some antigens in my case passed the filtering. Here is a subset of them: image

The last one for example has a foreigness_score of 0, so I though it should be removed (default parameter is foreignness_threshold = 1e-15). The min_DAI column also confuses me, because I though the value should be at least 10. Is for the filtering the mean_DAI considered?

Thank you in advance!

leeprichman commented 2 years ago

Hi Nickhir!

Sorry this is unclear, it also appears the function wasn't passing the arguments correctly. This function returns any mutant peptide with binding affinity ("Ensemble_score") < the affinity threshold (default 34). Then the other properties (dissimilarity, IEDB, and DAI) are used to label peptides in the "Recognition_Features" column but do not drop them from the table. So the only filtering that occurs is based on binding affinity < affinity_threshold.

Andrew is going to update the documentation to make it clearer and make sure the threshold arguments are passed correctly.

andrewrech commented 2 years ago

https://github.com/andrewrech/antigen.garnish/commit/551e332b519fa8cd6945f112bfec44bdd0de8019