A molecular 3D similarity method named SimG is presented which applies a downhill simplex method to compare the shapes and chemical features of a small molecule and a binding pocket (or ligand) with Gaussian volume overlap in order to identify the active compounds. The basic idea is quite similar to ROCS that the molecular similarity is evaluated by performing a Gaussian volume based alignment, but the proposed method is featured in two aspects. First, both structure-based and ligand-based strategies are supported. Second, a downhill simplex searching algorithm was then carried out to effectively evaluate the shape and chemical feature similarity between database molecules and the query binding site (or ligand) which could be used as a criterion to retrieve active molecules.