MTRXn records are not used when finding interfaces

[josemduarte]

Virus entries have MTRXn records needed to generate the entire AU (see http://www.wwpdb.org/documentation/format33/sect8.html#MTRIXn). Those are not taking into account when calculating all interfaces in crystal. E.g. 1auy

[josemduarte]

Note that in mmCIF format the matrices are in struct_ncs_oper.

Beware that it is also important to read the "iGiven" field (column 60 in MTRIXn in PDB format, field "struct_ncs_oper.code" in mmCIF). If iGiven=1 (or struct_ncs_oper.code="given") it means that the molecule is already given in the AU and thus that MTRIX record can be ignored. Normally this is used for identity operator but in some cases it is more of a "left-over" after having fixed the file, e.g. in 1d2q: the original file had only an A chain and a MTRIX record to construct chain B, later it was fixed and chain B was added, but the MTRIX record was left there with a "given" flag.

It is important to note then that the record is sometimes used in non-virus entries. See for instance 1a37. In that entry there are 2 missing chains (let's call them C,D) that are missing but that can be constructed through the MTRIX records. That is very wrong in my opinion.

See also this interesting post in pdb-l: https://lists.sdsc.edu/pipermail/pdb-l/2007-October/004131.html

And also this interesting newsletter from the phenix people (they even find the 1a37 case, plus many others): http://www.phenix-online.org/newsletter/CCN_2014_01.pdf