- Completely ignore the flexible parts by throwing them away (this is the wrong solution).
- Keep all atoms in the PSF and DCD files, but for the superposition use only atoms from the conserved structural core (this is the correct procedure).
The correct procedure can be implemented with xplor, vmd or carma. The carma way is to produce an ASCII file containing the indeces of the atoms that should be used for the superposition (much like the index file used for catdcd). The atom indeces are zero-offset (start from zero). Give this file the name fit.index. Then run carma with the flag "-index" :
carma -v -fit -ind CA.dcd CA.psf
The rms deviations reported by carma will be those calculated for the selected atoms alone (and not the whole molecule).