[Question]: Best way to handle Inf when comparing vectors?

EvanKomp commented 1 month ago

Feature Title

Best way to handle Inf when comparing vectors?

Feature Summary

Comparing smoothed matrices is hindered by Inf, not sure how best to handle them. Thanks for your tool!!!

Motivation and Context

I have run SURFMAP for a number of homologs, and have smoothed matrices for each.

I want to compare the matrices over different groups, eg. take the mean of a couple etc.

I am not sure what to do with Inf positions, maybe from a lack of understanding of what they actually are. Are these positions on the projection sphere without any nearby residuess to map?

Regardless, I am thinking and converting all Inf to Nan and doing Nan-ignoring operations over the vectors. Any wisdom is appreciated.

Technical Details and Implementation

No response

Duplicate Check

[X] I have searched the existing feature requests and did not find a duplicate.

EvanKomp commented 1 month ago

It seems as if any index that is Inf in one is Inf in all - though confirmation would be lovely

nchenche commented 1 month ago

Dear EvanKomp,

'Inf' values refer to cells in the border of the projection, id est cells that do not correspond to any protein residue(s). So indeed "Inf" value should be the same between all your maps/matrices.

To answer your question clearly: you should not consider "Inf" values when quantitatively operating on SURFMAP matrices.

Sorry for the delay, hope this will help.

Bests

i2bc / SURFMAP