This paper reviews the main efforts undertaken up to date in order to understand, rationalize and apply the similarity principle (similar compounds=>similar properties) as a computational tool in modern drug discovery. The best suited mathematical expression of this classical working hypothesis of medicinal chemistry needs to be carefully chosen (out of the virtually infinite possible implementations in terms of molecular descriptors and molecular similarity metrics), in order to achieve an optimal validation of the hypothesis that molecules that are neighbors in the Structural Space will also display similar properties. This overview will show why no single "absolute" measure of molecular similarity can be conceived, and why molecular similarity scores should be considered tunable tools that need to be adapted to each problem to solve.