Pushing the boundaries of computational approaches: special focus issue on computational chemistry and computer-aided drug discovery.

Abstract

This editorial introduces a new special focus issue dedicated to computational chemistry and computer-aided drug discovery to be published in the third quarter of 2016, for which the author will serve as a guest editor. To position this special issue some background information is provided and a few important aspects are highlighted. Medicinal chemistry publications frequently contain computational studies. Many of these studies are descriptive in nature. For example, a computational model is added to an experimental structure–activity relationship (SAR) investigation. This typically leads to the formulation of hypotheses as to how a compound might interact with its target or why active compounds differ in their potency. Another popular exercise is virtual screening aiming to identify novel active compounds using structureand/or ligand-based computational approaches. Less frequent are computational investigations reporting the prospective design of new compounds for synthesis and biological evaluation. In addition, compound property analyses are carried out, for example, to rationalize or predict drug-likeness or ADME characteristics. Furthermore, new computational concepts or methods are published to aid in drug discovery efforts at different levels. Computational studies reported in the medicinal chemistry literature are often characterized by a high degree of scientific heterogeneity. For example, modeling of compound binding modes is a popular exercise, often carried out to rationalize SARs. However, such predictions are frequently over-interpreted and putative interactions are discussed at the atomic level of detail as if they were experimental observations. On the other hand, there are also careful studies that evaluate putative binding modes taking the accuracy limitations of modeling into account and formulate experimentally testable hypotheses. However, the situation becomes particularly delicate when ‘multihypothetical’ strategies are pursued. This is illustrated by considering other examples from structure-based modeling. It is not uncommon that homology models are built for targets of interest, active compounds are docked into modeled binding sites, binding (free) energies are calculated for these ‘double-hypothetical’ ligand–target complexes, and correlated with experimentally determined affinities. What can one conclude from such exercises if close correlation between computed binding energies and experimental data is ultimately reported? Do such findings ‘validate’ the computational approach? Or might they represent the ‘luck of the draw’? Scientific views might well differ in such cases. Regardless, these types of investigations more or less follow a ‘look what computational methods can do!’ theme and often give incorrect impressions, especially to non-experts. This is far from being helpful for the further development of the field at Pushing the boundaries of computational approaches: special focus issue on computational chemistry and computer-aided drug discovery