I tried running this with several tweaks and iterations, I do not see diverse sequences being generated though. The sequence recovery rate is good, but almost all the sequences generated for a backbone are the same. Unlike something like ProteinMPNN where we can sample N number of novel sequences for a backbone. What am I missing here?
I tried running this with several tweaks and iterations, I do not see diverse sequences being generated though. The sequence recovery rate is good, but almost all the sequences generated for a backbone are the same. Unlike something like ProteinMPNN where we can sample N number of novel sequences for a backbone. What am I missing here?