 |
Center for Nonlinear Studies |
Intrinsically disordered proteins push the emerging view that proteins are best described as structural ensembles to the extreme, begging the question of how the free-energy landscape of these proteins differs from natively folded proteins. In particular, the dimensionality of the motion exhibited by natively folded proteins is predicted to decrease as the protein folds to its native structure, while dimensionality should stay relatively high for intrinsically disordered proteins due to their persistent flexibility. Dimensionality estimation techniques applied to molecular dynamics simulations have the potential to elucidate dynamical differences between various disordered proteins, allowing for their classification and categorization analogous to the structure-based methods currently used for classifying natively folded proteins. However, it is not clear that current dimensionality estimation techniques can actually recover the intrinsic dimensionality from simulations due to sampling limitations and the presence of thermal noise/fluctuations. In order to test their efficacy, we have developed a validation framework which utilizes a series of novel polymer models that exhibit dynamics characteristic of protein simulations, but are of known dimensionality by construction. Dimensionality estimation results from the polymer models indicate that (1) under/over estimation due to sampling/noise can be surprisingly predictable for some dynamical transitions, (2) while the estimators often fail to recover the true intrinsic dimensionality, ranking of relative estimates is still accurate, and (3) the two previous results can be leveraged to guide interpretation of results obtained from simulations of disordered and folding proteins.
Host: Kipton Barros, T-4 and CNLS