I had not done anything with par.lib, but I followed your suggestion and
edited it, removing things I wasn't using. I got rid of all the heme and
nucleic acid parameters, and this finally got rid of the MBNDPF error,
only to replace it with another concerning MANGPF (see below)! Any more
ideas?
Well, it's the same thing again: while MBNDPF is the maximum number of
bonds in the parameter file, MANGPF is the equivalent for angles. The
solution is the same - you just need to attack the ANGLES parameters in
par.lib this time. (There's also MDIHPF and MIMPPF, for dihedrals and
impropers.)