OPLSAA-M torsional parameters in oplsaam.prm file.

[BernardUPSaclay]

I would like to use torsion parameters for the peptide backbone to model PBLG (poly gamma-benzyl L-glutamate). I compare data in the Robertson-2015-JCTC paper and the oplsaa.prm, oplsaal.prm and oplsaam.prm files of tinker 8.10.5 distribution. Although oplsaa and oplsaal parameters correspond to those in Robertson paper, I find that parameters given in oplsaam.prm are different from those given in Robertson-2015 but are identical to those in oplsaal.prm.

E.g. the angle phi (C-N-Calpha-C) along the peptide backbone is described by the following atom CLASS (classes are different for oplsaa and oplsaa-l/m) :

file	    CLASS          V1/V2/V3                    V1/V2/V3
                          Robertson                   .prm files
-------------------------------------------------------------------------------
oplsaa    3-24-13-3    -2.365 / 0.912 / -0.85       -2.365 / 0.912 / -0.85
oplsaal  21-23-1-21    -0.596 / 0.279 / -4.913      -0.596 / 0.279 / -4.913
oplsaam  21-23-1-21    -2.511 / 0.21  / -0.200      -0.596 / 0.279 / -4.913

I find the same behavior for other angles along the peptide backbone.

Did I do some stupid mistake or misunderstood something in Robertson paper or tinker files organization? Or is there something wrong in the files?