Substrate specificity is a key feature of enzymes determining their applicability in biomaterials and biotechnologies. Experimental testing of activities with novel substrates is a time-consuming and inefficient process, typically resulting in many failures. Here, we present an experimentally validated in silico method for the discovery of novel substrates of enzymes with a known reaction mechanism. The method was developed for a model system of biotechnologically relevant enzymes, haloalkane dehalogenases. On the basis of the parametrization of six different haloalkane dehalogenases with 30 halogenated substrates, mechanism-based geometric criteria for reactivity approximation were defined. These criteria were subsequently applied to the previously experimentally uncharacterized haloalkane dehalogenase DmmA. The enzyme was computationally screened against 41,366 compounds, yielding 548 structurally unique compounds as potential substrates. Eight out of 16 experimentally tested top-ranking compounds were active with DmmA, indicating a 50% success rate for the prediction of substrates. The remaining eight compounds were able to bind to the active site and inhibit enzymatic activity. These results confirmed good applicability of the method for prioritizing active compounds-true substrates and binders-for experimental testing. All validated substrates were large compounds often containing polyaromatic moieties, which have never before been considered as potential substrates for this enzyme family. Whereas four of these novel substrates were specific to DmmA, two substrates showed activity with three other tested haloalkane dehalogenases, i.e., DhaA, DbjA, and LinB. Additional validation of the developed screening strategy with the data set of over 200 known substrates of Candida antarctica lipase B confirmed its applicability for the identification of novel substrates of other biotechnologically relevant enzymes with an available tertiary structure and known reaction mechanism.