Time-averaged restraints in molecular dynamics simulations offer a mea
ns to account for the averaging that is implicit in NMR spectroscopic
data. We present a systematic investigation of the parameters which ch
aracterise time-averaged distance restraints. Using previously publish
ed data for a small protein, chymotrypsin inhibitor 2, we identify con
ditions which can lead to undesirable heating or which grossly distort
the dynamics of the system.