free energy calc

Dear CCLers,
 We have been trying to "predict" the free energy change
 for replacing a neutral inhibitor bound to a protein
 into an isosteric one with a -1 charge via a
 thermodynamic cycle. Our approach consists of MD
 simulations with and explicit waters (TIP3P). We have
 tried both FEP and TI. We also experimented with
 particle mesh Ewald techniques. The bottom line is
 that the calculations are off by 2.5 to 5 kcal/mol
 > from reality depending on the protocol (we used
 the AMBER6 package). The problem probably does
 not reside in proper sampling as the simulations
 reproduce the conformational change seen between
 the 2 complexes (we have high resolution X-ray structures).
 Has anyone had any success with such type of simulations,
 i.e. cases involving a net change in charge? If so,
 could you provide us with recommended protocols.
 Christophe L.M.J. Verlinde
 Biomolecular Structure Center
 University of Washington - Box 357742
 Seattle, WA 98195
 T:(206)543-8865 F:(206)685-7002