Numerous methods are available for use as part of a virtual screening strategy but, as yet, no single method is able to guarantee both a level of confidence comparable to experimental screening and a level of computing efficiency that could drastically cut the costs of early phase drug discovery campaigns. Here, we present VSM-G (virtual screening manager for computational grids), a virtual screening platform that combines several structure-based drug design tools. VSM-G aims to be as user-friendly as possible while retaining enough flexibility to accommodate other in silico techniques as they are developed. In order to illustrate VSM-G concepts, we present a proof-of-concept study of a fast geometrical matching method based on spherical harmonics expansions surfaces. This technique is implemented in VSM-G as the first module of a multiple-step sequence tailored for high-throughput experiments. We show that, using this protocol, notable enrichment of the input molecular database can be achieved against a specific target, here the liver-X nuclear receptor. The benefits, limitations and applicability of the VSM-G approach are discussed. Possible improvements of both the geometrical matching technique and its implementation within VSM-G are suggested.