Efficient Partition Function Estimation in Computational Protein Design: Probabalistic Guarantees and Characterization of a Novel Algorithm

Abstract

By computational protein design we mean the use of computer algorithms to design new proteins or redesign existing ones. A significant challenge in this field involves computing the partition function of the ensemble of conformations that a protein can adopt. Due to the exponentially large number of possible states, there are too many conformations to explicitly count. One solution is to employ a probabilistic algorithm to estimate the number of conformations instead. In this work we implemented such an algorithm, studied its mathematical guarantees and analyzed its properties. Additionally we proposed different approaches to improve the convergence of the algorithm.