Have you looked at Gene3D and Superfamily? Both are based on PDB.
Blurb on Gene3D
[?]The Gene3D database is a large collection of CATH protein domain assignments for ENSEMBL genomes and Uniprot sequences Hide.
Proteins are generally composed of one or more functional regions, commonly termed domains. Domain annotation resources such as CATH provide curated domain assignments and a grouping of domains into families sharing evidence of homology (superfamilies). Although not a one to one relationship, a given domain superfamily will have a limited number of functions that its members fulfill in the whole-protein context. The combination of domains from different superfamilies gives rise to novel protein functions. Therefore, identifying the domains in a protein can give useful insights into protein function.
Finer grained groupings of superfamilies based on function (FunFams), and structure (SCs) are also provided. Gene3D is also unique in that it provides example of nested domains, where a given domain may have been inserted into another during evolution.
Blurb on Superfamily
SUPERFAMILY is a database of structural and functional annotation for all proteins and genomes.
The SUPERFAMILY annotation is based on a collection of hidden Markov models, which represent structural protein domains at the SCOP superfamily level. A superfamily groups together domains which have an evolutionary relationship. The annotation is produced by scanning protein sequences from over 1,700 completely sequenced genomes against the hidden Markov models.[?]
Gene3D is based on CATH
[?] CATH is a manually curated classification of protein domain structures. Each protein has been chopped into structural domains and assigned into homologous superfamilies (groups of domains that are related by evolution). This classification procedure uses a combination of automated and manual techniques which include computational algorithms, empirical and statistical evidence, literature review and expert analysis. [?]
Superfamily is based on SCOP
[?] The SCOP database, created by manual inspection and abetted by a battery of automated methods, aims to provide a detailed and comprehensive description of the structural and evolutionary relationships between all proteins whose structure is known. As such, it provides a broad survey of all known protein folds, detailed information about the close relatives of any particular protein, and a framework for future research and classification.[?]