The significance of significance: finding meaning in molecular similarity Molecular similarity, whether calculated from 1D, 2D or 3D properties, is one of the most widely used techniques in computational lead discovery. However, in spite of its wide use, the basic properties of molecular similarity are not well understood. This presentation will examine the distributions of scores from a variety of approaches to molecular similarity to determine how best to treat them in order to maximize the recovery of useful information. Appropriate transformations and processing for similarity scores along with rank-score plots and other representations of data from lead discovery will be investigated to determine how to minimize the false discovery rate (FDR) and maximize the effectiveness of data fusion in ligand-based virtual screening. These methods aim to maximize the impact of molecular similarity on projects in which it is used.