The PDB archive contains comprehensive descriptions of structural models coming from crystallography, NMR, and 3DEM. Each archival entry is denoted by a 4-character PDB identifier (e.g., 1VTL). In addition to atomic coordinates, details regarding the chemistry of biopolymers and any bound small molecules are archived, as are metadata describing biopolymer sequence, sample composition and preparation, experimental procedures, data-processing methods/software/statistics, structure determination/refinement procedures and statistics, and certain structural features, such as the secondary and quaternary structure. Primary experimental data coming from crystallography (structure-factor amplitudes or intensities) and NMR (restraints and chemical shifts) must be archived in the PDB. Voluntary archiving of diffraction images is currently supported by two resources that operate independently of the PDB, including the Integrated Resource for Reproducibility in Macromolecular Crystallography (IRRMC; www.proteindiffraction.org ) and the Structural Biology Data Grid Consortium (SBGrid; sbgrid.org [31 (link)]) both of which use digital object identifiers to make the data readily accessible. In addition, some synchrotron radiation facilities now store diffraction images in locally maintained repositories, with data retention and dissemination policies determined by the facility. BMRB [32 ] has long served as a public repository for NMR experimental data that are not stored in the PDB. Mass density maps used to derive structural models from 3DEM can be archived in EMDB [33 (link)]. Voluntary archival deposition of raw 3DEM images is currently supported by EMPIAR [34 (link)].
The first data format used by the PDB archive was established in the early 1970s, and was based on the 80-column Hollerith format used for punched cards [35 (link)]. Atom records included atom name, residue name, polymer chain identifier, and polymer sequence number. A set of “header records” contained limited metadata. The community readily accepted this format, because it was simple and both human- and machine-readable. However, the format also had limitations that became serious liabilities as structural biologists took the field to new heights. Structural models were limited to 99,999 atoms and relationships among various data items were implicit. These and other weaknesses of the legacy PDB format meant that deep subject matter expertise was required to both create and use software relying on this format. In the 1990s, the International Union of Crystallography (IUCr) charged a committee with creating a more informative and extensible data model for the PDB archive.
In response to the IUCR committee report, the Macromolecular Crystallographic Information File (mmCIF) was proposed [36 ]. mmCIF is a self-defining format in which every data item has attributes describing its features, including explicit definitions of relationships among data items. Most important, mmCIF has no limitations with respect to the size of the structural model to be archived. In addition, the mmCIF dictionary and mmCIF format data files are fully machine-readable, and no domain knowledge is required to read the files. At inception, the mmCIF dictionary contained over 3000 data items pertaining to crystallography. Over time, data items specific to NMR and 3DEM were added, and the dictionary was subsequently rebranded PDBx/mmCIF [37 ]. In 2007, it was decided that PDBx would be the PDB Master Format for data collected by the wwPDB. In 2011, major crystallographic structure determination software developers agreed to adopt this data model so that going forward all output from their programs would be available in PDBx/mmCIF.
In collaboration with community stakeholders serving on the PDBx/mmCIF Working Group (wwpdb.org/task/mmcif), the wwPDB continues to extend and enhance archival data representations. As of December 2014, PDBx/mmCIF became the official format for distribution of PDB entries. At the time of writing, the PDBx/mmCIF dictionary contained more than 4400 data items, including ~250 and ~1200 specific to NMR and 3DEM, respectively. PDBML, an XML format based on PDBx/mmCIF [38 (link)] and the requisite RDF (Resource Description Framework) conversion have also been developed to facilitate integration of structural biology data with other life sciences data resources [39 (link)]. Recently, XML and RDF-formatted BMRB data have been provided as BMRB/XML and BMRB/RDF, respectively [40 (link)], by which a federated SPARQL query linking the BMRB is made available to other databases. Finally, other structural biology communities are building on the PDBx/mmCIF framework to establish their own controlled vocabulary and specialist data items. For example, SASbDB has been working in collaboration with wwPDB partners to develop sasCIF [41 ], which builds on PDBx/mmCIF. In addition to accelerating development of the SASbDB archive, creation of sasCIF will allow for facile inter-operation with the PDB archive using a common exchange protocol based on PDBx/mmCIF.
In 1996, BMRB adopted NMR-STAR (a version of mmCIF) as its archival format [42 ]. As noted above, this format has been harmonized with PDBx/mmCIF and now serves as the preferred deposition format for NMR structures [43 ]. Historically, most NMR experimental data have been deposited in “native” format provided by each software package and archived “as is” in the PDB. Format harmonization was addressed in part by the NMR Restraints Grid, which can process restraint files and convert them to the NMR-STAR or CCPN formats [44 (link), 45 (link)]. In 2013 and 2014, community stakeholders participating in a pair of NMR format meetings convened by the wwPDB NMR VTF, recommended that an NMR Exchange Format (NEF) be developed for facile data transfer among NMR software packages and faithful conversion to NMR-STAR [46 (link)]. BMRB-led efforts are now underway to complete harmonization of NEF with NMR-STAR/PDBx/mmCIF to support NMR data deposition, annotation, and validation using the wwPDB unified global system (deposit.wwpdb.org).
Prior to 2015, reliance on the original PDB format made it necessary for large structure depositions (e.g., ribosomes/ribosomal subunits) archived in the PDB to be “split” into multiple entries, each with its own 4-character PDB identifier and legacy PDB-format file. This stopgap arrangement was entirely suboptimal. Splitting depositions among multiple PDB entries effectively precluded routine visualization of some of the most interesting structural models in the PDB archive, owing to software limitations. With adoption of the PDBx/mmCIF standard, every PDB archival entry is now stored as a single PDBx/mmCIF file, including 277 large structures that had previously been “split.” At the time of writing (and for the foreseeable future), archival entries are made available as a public service in “stripped down,” best-effort PDB legacy format files wherever possible. In time, visualization, computational chemistry, etc. software providers will need to adjust to the new format and use PDBx/mmCIF files directly.
The first data format used by the PDB archive was established in the early 1970s, and was based on the 80-column Hollerith format used for punched cards [35 (link)]. Atom records included atom name, residue name, polymer chain identifier, and polymer sequence number. A set of “header records” contained limited metadata. The community readily accepted this format, because it was simple and both human- and machine-readable. However, the format also had limitations that became serious liabilities as structural biologists took the field to new heights. Structural models were limited to 99,999 atoms and relationships among various data items were implicit. These and other weaknesses of the legacy PDB format meant that deep subject matter expertise was required to both create and use software relying on this format. In the 1990s, the International Union of Crystallography (IUCr) charged a committee with creating a more informative and extensible data model for the PDB archive.
In response to the IUCR committee report, the Macromolecular Crystallographic Information File (mmCIF) was proposed [36 ]. mmCIF is a self-defining format in which every data item has attributes describing its features, including explicit definitions of relationships among data items. Most important, mmCIF has no limitations with respect to the size of the structural model to be archived. In addition, the mmCIF dictionary and mmCIF format data files are fully machine-readable, and no domain knowledge is required to read the files. At inception, the mmCIF dictionary contained over 3000 data items pertaining to crystallography. Over time, data items specific to NMR and 3DEM were added, and the dictionary was subsequently rebranded PDBx/mmCIF [37 ]. In 2007, it was decided that PDBx would be the PDB Master Format for data collected by the wwPDB. In 2011, major crystallographic structure determination software developers agreed to adopt this data model so that going forward all output from their programs would be available in PDBx/mmCIF.
In collaboration with community stakeholders serving on the PDBx/mmCIF Working Group (wwpdb.org/task/mmcif), the wwPDB continues to extend and enhance archival data representations. As of December 2014, PDBx/mmCIF became the official format for distribution of PDB entries. At the time of writing, the PDBx/mmCIF dictionary contained more than 4400 data items, including ~250 and ~1200 specific to NMR and 3DEM, respectively. PDBML, an XML format based on PDBx/mmCIF [38 (link)] and the requisite RDF (Resource Description Framework) conversion have also been developed to facilitate integration of structural biology data with other life sciences data resources [39 (link)]. Recently, XML and RDF-formatted BMRB data have been provided as BMRB/XML and BMRB/RDF, respectively [40 (link)], by which a federated SPARQL query linking the BMRB is made available to other databases. Finally, other structural biology communities are building on the PDBx/mmCIF framework to establish their own controlled vocabulary and specialist data items. For example, SASbDB has been working in collaboration with wwPDB partners to develop sasCIF [41 ], which builds on PDBx/mmCIF. In addition to accelerating development of the SASbDB archive, creation of sasCIF will allow for facile inter-operation with the PDB archive using a common exchange protocol based on PDBx/mmCIF.
In 1996, BMRB adopted NMR-STAR (a version of mmCIF) as its archival format [42 ]. As noted above, this format has been harmonized with PDBx/mmCIF and now serves as the preferred deposition format for NMR structures [43 ]. Historically, most NMR experimental data have been deposited in “native” format provided by each software package and archived “as is” in the PDB. Format harmonization was addressed in part by the NMR Restraints Grid, which can process restraint files and convert them to the NMR-STAR or CCPN formats [44 (link), 45 (link)]. In 2013 and 2014, community stakeholders participating in a pair of NMR format meetings convened by the wwPDB NMR VTF, recommended that an NMR Exchange Format (NEF) be developed for facile data transfer among NMR software packages and faithful conversion to NMR-STAR [46 (link)]. BMRB-led efforts are now underway to complete harmonization of NEF with NMR-STAR/PDBx/mmCIF to support NMR data deposition, annotation, and validation using the wwPDB unified global system (deposit.wwpdb.org).
Prior to 2015, reliance on the original PDB format made it necessary for large structure depositions (e.g., ribosomes/ribosomal subunits) archived in the PDB to be “split” into multiple entries, each with its own 4-character PDB identifier and legacy PDB-format file. This stopgap arrangement was entirely suboptimal. Splitting depositions among multiple PDB entries effectively precluded routine visualization of some of the most interesting structural models in the PDB archive, owing to software limitations. With adoption of the PDBx/mmCIF standard, every PDB archival entry is now stored as a single PDBx/mmCIF file, including 277 large structures that had previously been “split.” At the time of writing (and for the foreseeable future), archival entries are made available as a public service in “stripped down,” best-effort PDB legacy format files wherever possible. In time, visualization, computational chemistry, etc. software providers will need to adjust to the new format and use PDBx/mmCIF files directly.
