Spaces:
Running
on
T4
Running
on
T4
SolubleMPNN bug?
#14
by
jvarga92
- opened
Hi,
I was trying to redesign a membrane protein (claudin as in the SolubleMPNN paper) by setting the model type to soluble. However, the sequences that I got were as hydrophpobic on the outside as before. Maybe there is a bug with it?
I just verified (you can too since the code is available under the files tab) and it is using the correct weights.
Perhaps you need to play with sampling temperature or add backbone noise?