Biomolecule Electrostatic Optimization with an Implicit Hessian

Computational rational drug design is the application of computer simulation techniques to improve screening processes for new drugs or to design them de novo. The goal is to identify molecules that have high affinity and specificity for a target molecule. Optimizing the electrostatic binding free energy is tractable under certain assumptions as a quadratic optimization problem, but computationally expensive simulations have traditionally been required to determine the associated Hessian. Prior work showed that significant performance gains could be achieved by coupling physical simulation directly to the optimization process and avoiding the calculation of the Hessian. The present paper details recent improvements to this method and the implementation of a practical, full-scale code. Computational results demonstrate the code’s efficiency on large problems and accuracy solving a realistic biomolecule optimization problem.