This paper describes a method for selecting a small, highly diverse su
bset from a large pool of molecules. The method has been employed in t
he design of combinatorial synthetic libraries for use in high-through
put screening for pharmaceutical lead generation. It computes diversit
y in terms of the main factors relevant to ligand-protein binding, nam
ely the three-dimensional arrangement of steric bulk and of polar func
tionalities and molecular entropy. The method was used to select a set
of 20 carboxylates suitable for use as side-chain precursors in a pol
yamine-based library. The method depends on estimates of various physi
cal-chemical parameters involved in ligand-protein binding; experiment
s examined the sensitivity of the method to these parameters. This pap
er compares the diversity of randomly and rationally selected side-cha
in sets; the results suggest that careful design of synthetic combinat
orial libraries may increase their effectiveness several-fold.