Conformational Ensembles from Experimental Data and Computer Simulations

Conformational Ensembles from Experimental Data and Computer Simulations

Poster Abstracts

90-POS Board 10 Guiding Stochastic Optimization Algorithms with Experimental Data to Model Protein Energy Landscapes and Structural Transitions

Amarda Shehu . n/a, Fairfax, USA.

Some of the most complex human disorders are driven by DNA mutations that percolate to protein dysfunction. While it is known that mutations percolate to dysfunction by changing the energy landscape and in turn the structural dynamics of a protein, quantifying changes to the landscape and dynamics of a protein in response to a mutation remains elusive. Even reconstructing the energy landscape of the healthy, wildtype form of a protein is currently out of reach. While the challenges to wet- and dry-laboratory techniques are different in nature, they all relate to the fact that the dynamics of interest, corresponding to structural transitions on the energy landscape, spans disparate spatio-temporal scales. Recent work in our laboratory is exploiting the wealth of accumulated structural data on a protein’s variants to address some of the outstanding challenges to in-silico models of equilibrium dynamics. Stochastic opimization algorithms are developed in our laboratory to build detailed, yet resource-aware maps of protein energy landscapes. These algorithms exploit experimental data to make informed algorithmic decisions such as variable selection and variation operators. The algorithms first construct unstructured, sample-based maps of a protein’s energy landscape, and then enrich such maps with connectivity information to obtain the connected landscape. The latter can provide information on any structural transitions of interest, as well as yield summary statistics on dynamics. Studies on several proteins show this approach is promising and can reconstruct landscapes that currently remain beyond the reach of molecular dynamics and monte carlo-based approaches. Results on specific proteins of importance to human disorders make the case that the computed, connected landscapes advance our understanding of the role of dynamics on how mutations percolate to dysfunction and even provide directions of relevance for novel therapeutics.

127