Evolutionary Search for Mapping Minima in the Protein Energy Surface
Ph.D. Defense Talk - Work by B. Olsong
and A. Shehu* at ACM Conf on Bioinf and
Comp Biol (BCB) 2013, AAAI-W Workshop on AI and Robotics Methods for
Computational Biology 2013, Genet Evol Comp Conf (GECCO) 2013, BMC
Struct Biol J 2013, IEEE BIBM-W Comp Struct Biol Workshop (CSBW)
2012, Proteome Sci J 2013, IEEE Intl Conf on Bioinf and Biomed
(BIBM) 2012, Proteome Sci J 2012, and IEEE Intl Conf on Bioinf and
Biomed (BIBM) 2011.
Protein energy surfaces are nonlinear and multimodal, which makes them suitable systems to study with evolutionary search/optimization algorithms. We are currently exploring such algorithms to effectively sample local minima in the protein energy surface. These minima are of relevance when studying thermodynamically-stable and semi-stable structural states that a native protein uses for its biological function or a variant employs for loss of function. Our focus is on equipping the basic algorithmic frameworks with domain-specific (biophysical) knowledge on proteins and then pursuing adaptations of the basic frameworks for an enhanced exploration capability. The large objective is to employ these algorithms to obtain a detailed characterization of the structure space and model the structure-function relationship in protein and protein-like systems.
Our work has investigated the basic Basin Hopping framework, more powerful hybrid population-based frameworks, implementation of various global and local moves in evolutionary search algorithms, and the incorporation of multi-objective optimization through Pareto-based metrics to attenuate the reliance on noisy energy functions and obtain a more diverse conformational ensemble. Details can be found in the related pages and publications.
On this Project:
-
Brian Olson
Sameh Saleh (Undergraduate Student)
Irina Hashmi
Kenneth De Jong
Amarda Shehu