Scoring the compatibility of a sequence with a given fold is a cornerstone task in computational protein design. Because vast sequence spaces are considered in design applications, scoring functions must be fast. On the other hand, as modern computer clusters become more ubiquitous and powerful, this speed requirement is continually softened. Unfortunately, this does not necessarily translate into better scores, as traditional scoring functions contain built-in assumptions and approximations that do not vanish at the infinite sampling limit. I propose a novel method, which can estimate the absolute free energy of a structural state using only information from a canonical sampling thereof (e.g. via Molecular-Dynamics or Monte-Carlo simulations). The method, dubbed VALOCIDY, is mathematically exact at the long sampling limit, whereas estimates based on limited amounts of simulation can be accurate enough to enable comparisons between very different structural states (e.g. folded and unfolded states or alternative folded conformations). I will briefly present the theoretical foundation of the approach and demonstrate several applications, including a small design problem. Tradeoffs between increased accuracy of VALOCIDY estimates, over traditional scoring, and added computational expense will need to be carefully examined, especially in larger systems, but the added physical interpretability is certainly a much needed property in the context of design applications.

Toward an ensemble-centric view of structure for protein design

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.