Using multiobjective optimization and energy minimization to design an isoform-selective ligand of the 14-3-3 protein

Computer simulation techniques are being used extensively in the pharmaceutical field to model protein-ligand and protein-protein interactions; however, few procedures have been established yet for the design of ligands from scratch ('de novo'). To improve upon the current state, in this work the problem of finding a peptide ligand was formulated as a bi-objective optimization problem and a state-of-the-art algorithm for evolutionary multiobjective optimization, namely SMS-EMOA, has been employed for exploring the search space. This algorithm is tailored to this problem class and used to produce a Pareto front in high-dimensional space, here consisting of 2322 or about 1030 possible solutions. From the knee point of the Pareto front we were able to select a ligand with preferential binding to the gamma versus the epsilon isoform of the Danio rerio (zebrafish) 14-3-3 protein. Despite the high-dimensional space the optimization algorithm is able to identify a 22-mer peptide ligand with a predicted difference in binding energy of 291 kcal/mol between the isoforms, showing that multiobjective optimization can be successfully employed in selective ligand design.