The term "generate" explained

CNS reads the coordinate (PDB) file twice:

 segment
   chain
     @@CNS_TOPPAR:protein.link
     coordinates @@file.pdb     <--- first read
   end
 end

 coordinates @@file.pdb    <--- second read

In the first pass, only the residue names are read, from which it generates an internal list of the atoms which should be present and how they are connected (by using the topology files in the library/toppar directory). After this internal molecular topology has been constructed the coordinates are read a second time. At this point the coordinate records must match the internal topology for the coordinates to be read in. If there are inconsistancies in the atom naming but not the residue naming in a coordinate file then the first pass will succeed (as the residue names are correct). However, the second stage where the coordinates themselves are read will fail for some atoms. The following message will be obtained for those atoms:

 %READC-ERR: atom 1HOE 74   LEU  OT   not found in molecular structure

If there are some atoms that are not present in the input coordinate file or are incorrectly named (compared to the CNS topology library) then there will be coordinates missing for some atoms. This will be reported after the coordinates have been read for the second time:

 %READC-WRN: still    170 missing coordinates (in selected subset)

Coordinates (X,Y,Z) for these atoms must either be generated or the atoms deleted. For crystallographic applications in general the missing atoms are hydrogens which are deleted prior to writing out the final molecular topology information.