Every protein structure in TMPL has a unique accession code, whose pattern is dependent on whether the entry is a theoretical or an experimental model.
The entry codes of native structures—and their martinized versions—include the PDB ID, followed by the representation ('AT' for atomic, or 'CG' for coarse-grained), and end by 'D' (default) or 'G' (geometric), which correspond to the methods used by the orientation algorithm to determine the membrane boundaries (Postic et al., 2016a). Thus, the pattern of the accession codes for native structures can be written as:
For predicted structures, the entry code successively includes the UniProtKB AC of the protein chain, the source of the model—i.e. 'S' (SWISS-MODEL), 'M' (ModBase), 'E' (EVfold_membrane), 'O' (ORION), or 'U' (user-deposited)—, the number of the model (from 1 to 99; when more than one model of the same protein), and the positioning method employed by ANVIL ('D' or 'G'). Thus, the pattern for predicted structures is:
Finally, when more than one valid orientations are available, a number is appended to the accession code.
Protein 3D structures in TMPL can be visualized on the website, through the WebGL-based viewer PV (Biasini, 2014).
- hold left click to move the structure;
- hover a residue to display its name, number and chain;
- double-click on any part of the structure to change the center of rotation;
- use the scroll wheel to zoom in and out.
Please note that larger structures may take longer to load.
Of course, all protein structure of TMPL can be individually downloaded to be visualized with any locally installed molecular viewer (PyMOL, VMD, Chimera, RasMol etc.).
A quick search can be performed by using the text area of the top-right corner. This will match with the name, organism, and PDB, Pfam or UniProt codes. For more specific queries, the 'Search' form shall be used.
The TMPL database allows users to deposit new membrane protein structures. It only requires:
- assigning the lipid bilayer planes with the OREMPRO web server;
- completing a form about the protein (name, organism etc.);
- and our manual verification of the new entry.
For protein structures that already exist in the database, users can still deposit alternative membrane assignments, resulting from the different tunings of the OREMPRO parameters—in particular, the membrane propensity scale on which the positioning algorithm is based (Postic et al., 2016b), and which may not be relevant for all membrane proteins (different membrane lipid compositions may require different hydrophobicity scales).
For each protein model in TMPL, a structural assessment of the transmembrane domain is provided in the form of a score (MAIDEN score: Postic et al., 2015). It is aimed at comparing the quality of different models of the same protein—i.e. the relative quality. The absolute quality is represented by a color code—from red for bad models, to blue for near-native structures—which is derived from a Z-score calculation of the MAIDEN pseudo-energy.
These measures actually reflect the quality of both the structural model and its orientation in the lipid bilayer. Thus, a high quality structure with a wrong membrane assignment that would include globular parts of the protein in the lipid bilayer boundaries, is expected to have a relatively high MAIDEN score with a yellow or red color tag.
The TMPL database is updated every time a new protein structure is manually deposited and validated. Besides these irregular additions, TMPL is automatically updated every 6 months, by searching for new models in the PDB, ModBase and SWISS-MODEL databases.
It must be noted that the number of entries in TMPL may sometimes decrease. This is due to the fact that some of the automatically added entries are not 'actual' transmembrane proteins, and are therefore removed from the database when detected. Most of the time, it corresponds either to structural predictions that only cover a short part of the template protein chain, or to models of globular parts of transmembrane proteins.