lfinder is only partially robust to mismatches between MOAD naming schemes and pdb structure lines

[lucagl]

The naming scheme of binding moad is to associate any residue name appearing on a line to a increasing res number, for each chain (for instance 1cqf--> BGC GAL GLA F:1 G:1 I:1 ..) will look for ligand1=(BGC F 1, GAL F 2, GLA F 3), ligand2 =(BGC G 1, GAL G 2, GLA G 3), ligand2 =(BGC I 1, GAL I 2, GLA I 3) ecc..

So far, the strategy used to overcome mismatches is to skip progressively in order (from the MOAD naming scheme) the name-chainID-ResNumber. This works most of times but not always. An example is 4z0u.pdb where the pdb skips the first 4 names and start from the 5th but keeping to index of the number at the beginning (usually everything is shifted accordingly).

In general, one could device more flexible strategies to overcome mismatches

[lucagl]

Other examples of failures due to mismatches: 3ebb, 4ii9, 4k76, 5azg