I have a large dataset of Uniprot-Accessions for proteins. If I look at a specific position in these protein's sequences, I want to know if this residue is buried or exposed (i.e. on the surface). These positions represent dbSNP entries that cause missense mutations, if that is of any help. The goal is to combine that with Pfam domain data to predict whether a residue is part of a protein-protein interaction interface.

I know that I could download the PDB files, run DSSP and calculate the relative surface area, but before implementing that I wanted to make sure there isn't any database with entries like this that I have missed.