You need an index groups that specifies those residues; (almost) all Gromacs analysis tools accept index groups.  All of the common ones do.

you better use AMBER for that and specify 45-187 in input file.

for Gromacs, g_rms -s md.tpr -f md_noPBC.xtc -o rmsd.xvg -tu ns

I think, as in most GROMACS tools, the -n option and an index file containing the indices of the atoms of your amino acids with the proper selection of groups should do. Be sure to choose and carry out an apropriate fitting method before doing the RMSD (e.g. fit this residue, backbone only or whatever fits your purpose the most).

I think that you can create groups of atoms (to define them in the index file with the tool make_ndx) and using the command:

g_rms -s .gro -f .xtc -n -fit rot+trans -o .xvg

Select group for least squares fit:

Select group for RMSD calculation:

Note: .gro -It is of the initial structure with ions and solvent

.xtc -Obtained with the tool trjconv when it finishes the phase of production (to account for any periodicity in the system).

thank you

You need an index groups that specifies those residues

RMSD is easily accomplished using one or other of the various molecular graphics tools. I would recommend YASARA which has several methods for doing this. My own preference among these is the Sheba superposition tool.