The drug designer´s guide to selectivity

Although several dual activity and so-called dirty drugs, which are active on more than one drug target or mediate more than one effect, have made it to the market the ruling paradigm in drug discovery today is to deliver selective drugs. However, it is not always a trivial task to make a priori predictions of the possibilities for achieving ligand selectivity for a given target, and compounds in late developmental phase, or even after launch, may still be found to possess activity and effects not previously foreseen. It would therefore be of great benefit if one could display any notions addressing selectivity based on database knowledge and simple calculations. In the present study, a sequence alignment-based approach will be presented and it will be applied to the PDZ class of protein domains. This protein class consists of small globular domains that easily superimpose using their 3D structure.In a recent human genomics investigation, a PDZ containing protein, i.e., DLG5, was shown to comprise a single nucleotide polymorphism related mutation in one of the DLG5 protein, which was found to be associated to an increased risk of having a Crohn's disease diagnosis. Consequently, it would be of medicinal interest to design ligands selective to the DLG5 PDZ domains and the present process aims at the possibility to achieve that using a novel chemometric approach, including a new set of amino acid property scales. Thus, in the present paper, selectivity aspects as seen from the protein target perspective will be outlined. This approach is relying on adequate multiple sequence alignments (MSA) and the PDZ protein domains provide a good example due to their firm homology in 3D structure.