Hierarchical experimental design exemplified by QSAR evaluation of a chemical library directed towards the melanocortin 4 receptor

A design strategy for how to select a smaller subset of compounds to synthesize, test for activity and finally establish a QSAR model was investigated. The biological target of interest was the melanocortin 4 receptor (MC4R), which is assumed to have an important role in the control of feeding behaviour and is therefore of great interest as a drug target. A subset of our in‐house compound collection was used to investigate and illustrate the proposed design strategy. The selected compound set has a phenyl core motif in common with five substituents in the ring, which could be varied independently and therefore represented a suitable design scaffold to investigate. By using statistical molecular design in combination with the proposed hierarchical design approach, a training set of compounds was selected. As design variables for the substituents, principal properties based on a set of 110 substituents characterized with σ, π and Verloop parameters were used. By using measured affinity data towards MC4R as a response variable together with the principal properties, a QSAR model was established which proved to have good predictive ability. Two important conclusions were that none of the compounds in the test set were predicted as false negatives by the QSAR model and that the proposed design approach appears to work well. The strategy is implemented in an ongoing project, and the usefulness of hierarchical design of experiments will be explored further.