Computational ligand design by free energy minimization

We present a computational method for de novo ligand design based on simulated annealing optimization of an empirical free energy function which has been derived from regression analysis of protein-ligand complexes. Structures are built by joining together molecular fragments from the Available Chemicals Directory. The algorithm navigates through the space of molecular structures, conformations and rigid-body transformations. The diversity of the generated structures can range from completely de novo to products of a combinatorial synthetic reaction scheme and docking of single molecules. The synthetic feasibility of the proposed molecules is considered by the incorporation of penalty terms to reduce undesirable connectivity and chemical features. Test results for tripsin are presented.