The application of ligand-based virtual screening (LBVS) is commonly extended in the screening of vast chemical libraries to find novel chemical matter with respect to a query compound of your interest using different molecular descriptions. Different to 1D or 2D descriptors, when performing 3D-LBVS, the selection of a proper compound conformation will have a great impact on the results obtained.

In most cases, the selection of the bioactive conformation (that is, the conformation of a compound in the binding pocket of a protein) of the query molecule (the molecule that will be used as template to find new similar compounds) may be the most recommended choice. However, the availability of these conformations is affected by the availability of structural information derived from crystallographic (X-Ray) or nucleic magnetic resonance (NMR) studies. Even if a complex ligand-protein is available, there are many aspects to consider the conformation as a starting point, such us the method applied to co-crystalize the complex or the resolution of the crystal, in the case of X-Ray studies.

Recently, we have published a paper in collaboration with the Computational Biology Chemistry & Gastronomy of the University of Barcelona with the goal of evaluating the impact of the selection of different methods to determine the conformation of the query molecule. The study entitled “On the relevance of query definition in the performance of 3D ligand-based virtual screening” pays also special attention to how not only the conformation of the query, but also the similarity of the screened compounds with respect to the query may impact the criteria to select a method to determine the conformation of the query.

The study compares five different methods: the X-Ray structure, the energy-minimized X-Ray structure, the lowest energy conformer sampled for the free compound in the gas phase, the lowest energy conformer of the free compound in water and an ensemble of accessible conformers in the gas phase. To compare the five different starting points, two 3D-LBVS tools were applied: PharmScreen, which takes advantage of 3D molecular field descriptors derived from quantum mechanics calculations to compare the similarity between two molecules; and Phase Shape, is a pharmacophore/volume overlap modeling solution where each ligand is represented by a set of points that account for specific chemical features in 3D space.

Two different libraries were evaluated for this study: the DUD-E+ and DUD-E+-Diverse, a Morgan Fingerprint-filtered version that minimizes the 2D resemblance observed for DUD-E between the query and the active compounds.

The results and conclusions of the study are explained extensively in the article. However, three different results are worth highlighting:

• The choice of the query molecule conformation has a big impact on the identification of novel compounds, moreover when our goal is to identify more chemical diversity. In the case of the present study, this affects the recovery of active compounds that have more chemical diversity with respect to the query molecule.
• The selection of the energy-minimized conformation of the query molecule could be an advantage in the identification of compounds with similar scaffolds since the conformational space adopted both for query and screened similar compounds will be more similar due to the chemical resemblance.
• Using X-Ray conformation will be a better choice when the conformation adopted by the query is influenced by the topology or specific interactions in the binding pocket.