A novel program for the superposition of flexible molecules, slate, is pres
ented. It uses simulated annealing to minimise the difference between the d
istance matrices calculated from the hydrogen-bonding and aromatic-ring pro
perties of two ligands. A method for generating a molecular stack using mul
tiple pairwise matches is illustrated. These stacks are used by the program
doh to predict the relative positions of receptor atoms that could form hy
drogen bonds to two or more ligands in the dataset. The methodology has bee
n applied to ligands binding to dihydrofolate reductase, thermolysin, H-3 h
istamine receptors, alpha (2) adrenoceptors and 5-HT1D receptors. When ther
e are sufficient numbers and diversity of molecules in the dataset, the pre
diction of receptor-atom positions is applicable to compound design.