The TMDET algorithm and the generation of PDBTM database

The heart of PDBTM database is the TMDET algorithm, which is designed to detect the location of the transmembrane (TM) region within protein structures represented in the PDB. TMDET operates by analyzing solvent accessibility on hydrophobic and amphipathic residues and other structural factors.

The original coordinates of the molecule are transformed so that the Z-axis coincides with the membrane normal and new origo is along this axis at the membrane half-width. The transformation matrix given in the xml/json files is the same form as matrix records in the PDB (e.g. REMARKS 350). The first three column give the 3x3 rotational matrix and the fourth coloumn describes the translation vector.

The PDBTM database is updated every week. The update includes a combined process of automatic and manual steps. First we use TMDET to detect candidate membrane proteins from PDB and to define the most likely position of the lipid bilayer relative the the protein. Although these steps are automatic, for each entry we manually validate the selection of the protein and the membrane position. We might overrule the raw output to either remove false positives, or to better reconstruct the membrane bilayer.