In order to compare SSM and another protein 3D matching service, we first obtain results from that service for a particular PDB entry. A full PDB set and widest ranges for RMSDs, P-values, Z-scores etc. are specified where applicable. The results are coming either as HTML pages or generated e-mails (DALI), which need to be parsed in order to extract the data.

Then SSM is run for the same PDB entry, with the lowest acceptible match percent set to 10%. That normally results in a much larger number of matches than obtainable from other services. Most of them, naturally, have a very low significance.

The obtained from SSM and other services results are corresponded by PDB code and chain ID (except for DEJAVU results), and ordered by decreasing the length of Ca-alignment given by SSM; the non-corresponded SSM results are filtered out.

The problem with DEJAVU is that it does not give chain IDs of its matches. We assume that, in case of multiple-chain (target) PDB entries, DEJAVU reports the best match found. We therefore correspond it with the best match for that particular target protein obtained from SSM. This implies that SSM results are corresponded with DEJAVU by PDB code but not necessarily by chain ID.

While the results from SSM, VAST, DALI and DEJAVU were obtained by means of actual calculations, CE offers ready tables of structural neighbours. These tables do not include the query structure, therefore CE results do not necessarily start with full alignment length and zero RMSD.