From finance to molecular modeling algorithms: The risk and return heuristic

While machine learning techniques have greatly increased molecular modeling capabilities, the frequent reliance on stochastic algorithms is a limiting factor due to slow optimization processes. Faster algorithms are thus sought for large scope projects. Nevertheless, stochastic algorithms also provide a distinct advantage by providing a solution ensemble rather than a single optimal solution. Producing a large ensemble of solutions is critical in problems where not all solution aspects are predictable and unpredictable properties may be key to ultimate success. Similar problems have been tackled previously before the advent of machine learning, in the field of finance. In 1952, Harry Markowitz introduced the modern portfolio theory (MPT), which uses the heuristics of risk and return to optimize a financial portfolio. In this study we will introduce an implementation of MPT heuristics in the field of protein-protein and peptide-protein interface design and show examples of its usage.