2D Autocorrelation behavior #3790
Unanswered
edwardclem
asked this question in
Q&A
Replies: 1 comment
-
Hi @edwardclem, |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi everyone!
I've been testing some of the descriptors in
rdkit.Chem.Descriptors
and found a behavior that may or may not be a bug - for some short carbon chains (CCCC
for example) some of the 2D autocorrelation descriptors report nans.My understanding of these descriptors is that they represent autocorrelations between pairs of atoms at specified distances, using some atomic property as a weighting function (6 properties are specified in the code, so I'm assuming the 192-dimensional vector refers to these 6 weights at distances up to 24 bonds?). It makes sense that some of these descriptors wouldn't be applicable to this molecule, as there aren't any paths longer than 4, but I'm not sure why they would report nans unless there's some normalization issue happening somewhere. Is this an expected behavior?
Steps to reproduce:
I've experienced this using rdkit 2020.09.03 on Ubuntu 18.04 and macOS 10.14. The same thing also appears when calling the individual autocorrelation descriptor functions instead of the
CalcAUTOCORR2D
function.Beta Was this translation helpful? Give feedback.
All reactions