MMDB - Entrez's Structure Database

NCBI's Entrez includes a database of experimentally determined three-dimensional biomolecular structures. Most 3D-structure data are obtained from X-ray crystallography and NMR-spectroscopy; they provide a wealth of information on the biological function, on mechanisms linked to the function, and on the evolutionary history of and relationships between macromolecules. Our goals in adding structure data to Entrez are to make this information easily accessible to biologists, and to facilitate comparative analysis involving 3-D structure.

NCBI's structure database is called MMDB (Molecular Modeling DataBase), and it is a subset of three-dimensional structures obtained from the Protein Data Bank (PDB), excluding theoretical models. MMDB is a database of ASN.1-formatted records. It was designed for flexibility, and as such, is capable of archiving conventional structural data as well as future descriptions of biomolecules, such as those generated by electron microscopy (surface models).

The structure database is considerably smaller than Entrez's protein or nucleotide databases, but a large fraction of all known protein sequences have homologs in this set, and one may often learn more about a protein by examining 3-D structures of its homologs. Protein sequences from MMDB are extracted and available in the Entrez protein sequence database. They are linked to the 3-D structures, therefore it is possible to determine whether a protein sequence in Entrez has homologs amongst known structures by examining its Related Sequences or Protein Neighbors and checking whether this set has any Structure Links.

How and when is MMDB updated?

The Protein Data Bank (PDB) is updated weekly, and major releases (also available on CD-Rom) are issued several times a year. The NCBI mirrors PDB for the purpose of making MMDB. Once a week, the new entries are parsed and loaded into the PKB (Protein Knowledge Base) database.

The software which builds the PKB database has a comprehensive PDB parser. This parser is able to extract the sequences and citations from PDB records, in addition to the structural information. All these data elements are translated into ASN.1. An explicit chemical graph representing the connectivity of the structure based on its chemistry is created in PDB and written out into ASN.1 records. PDB data are checked for consistency in the purported chemistry, the sequence, and the 3-D coordinates.

Entrez users may occasionally notice that the sequence of a PDB-derived entry differs slightly from the PDB file, since all non-standard or chemically modified residues, as judged by their 3-D structure, are explicitly identified as such in MMDB. These changes to PDB data are intended to support computational applications such as homology modeling and structure comparison.

The original models (coordinates) provided by PDB are preserved and translated into ASN.1. In addition, new models representing single-coordinate per-atom molecules (i.e any multiple conformations for single atoms are removed) and alpha-carbon backbones are created. PKB also generates explicit bonding and secondary structure information in addition to any provided by PDB, and this information is used, for example, to create vector models for the purposes of structural comparison and alignment.

With each update the taxonomy is assigned to each chain of the new entries by an automated method. When new structure data are added to the Entrez-system, they get cross-indexed to nodes in the taxonomy hierarchy, to citations in PubMed, and to sequences derived from the structures.

With each update, structure neighbors for all the protein chains and their individual domains are calculated by comparing the three-dimensional architectures to all the other previously determined structures. This set of relationships can be used to search for Structure Neighbors from the structure summary pages of each entry as well as from within Entrez. All of this makes MMDB a "value-added" structure database. In Entrez, powerful search and selection mechanisms are available. Structure data can be retrieved by organism (or higher node in the taxonomical hierarchy), by publication date, author names, or substance names, just to name a few.

The structure neighbor data are an invaluable tool in studying remote homologs and distant evolutionary relationships. Further software development will focus on easy accessibility to this wealth of biological data.

The structure database may be queried directly, using specific fields such as author names, or text terms occurring anywhere in the structure description. Entry points for queries are the Search Bar at the top of all Structure Group WWW pages or the WWW-Entrez interface to the 3-D structure database.

Alternatively you can use a PDB 4-character code or a numerical MMDB-Id to retrieve structure summary pages directly: PDB/MMDB Code(s):