Mapping from Sequences to Domains/Scaffolds

[brucejwittmann]

I'm working with Dataset 3 and am trying to compare K50 values for proteins that share a fold. I can't see an easy to way to identify which proteins go to which domain (or fold) from the columns in "K50_Dataset3.csv", however. Do you happen to have a table available that maps the name of each entry in K50_Dataset3.csv to its associated domain? In all, I'm hoping to be able to distinguish (1) which natural proteins share a wild type and (2) which designed proteins share the same base scaffold.

[JinyuanSun]

Hi, have you find a way to solve this?

[brucejwittmann]

Somewhat, though it's definitely not perfect. I approached it two separate ways:

1. I made the assumption that all entries with the same pdb name prior to the mutation information were from the same group. This works well for about 90% of the entries, and I could confirm by clustering with mmseqs that my initial assumption about the pdb name was good.
2. I used mmseqs easy-cluster (see here) to cluster at 90% sequence identity, 80% coverage. I assumed that anything that shared a cluster was from the same domain. This is obviously imperfect, but it at least groups similar proteins.