Using a 4D-QSAR approach (software Quasar) allowing for multiple-conformati
on, orientation and protonation-state ligand representation as well as for
the simulation of induced-fit phenomena, we have validated a family of rece
ptor surrogates for the 5-HT2A receptor system. The evolution was based on
a population of 200 receptor models and simulated during 6,000 cross-over s
teps, corresponding to 30 generations. It yielded a cross-validated r(2) of
0.951 for the 23 ligands of the training set and a predictive r(2) of 0.85
9 for the 7 ligands of the test set. In this simulation, all ligand molecul
es were represented by four different conformers, obtained from a Monte-Car
lo search in implicit aqueous solution. A series of six scramble tests (wit
h an average predictive r(2) of -1.05) indicate a high sensitivity of the s
urrogate family towards the biological data.
The quantitative analysis of the contribution of the individual functional
groups to the free energy of ligand binding, DeltaG degrees, reveals that t
he key factors for strong binding-and hence activity-are the ligand desolva
tion energy and the costs associated with induced fit, the adaptation of th
e receptor-binding site to the ligand topology. While the ammonium function
ality is essential for recognition, its contribution to DeltaG degrees is n
ot favorable due to a high desolvation energy; important groups are rather
methoxy and halide substituents. For most ligand molecules, the evolution d
oes not select the lowest-energy conformer, contrary to previous assumption
s in a corresponding 3D-QSAR study.