Browsing by Author "Tekpinar, M."

Now showing 1 - 2 of 2

Comparative Investigation of Normal Modes and Molecular Dynamics of Hepatitis C Ns5b Protein
(Iop Publishing Ltd, 2016) Asafi, M. S.; Yildirim, A.; Tekpinar, M.
Understanding dynamics of proteins has many practical implications in terms of finding a cure for many protein related diseases. Normal mode analysis and molecular dynamics methods are widely used physics-based computational methods for investigating dynamics of proteins. In this work, we studied dynamics of Hepatitis C NS5B protein with molecular dynamics and normal mode analysis. Principal components obtained from a 100 nanoseconds molecular dynamics simulation show good overlaps with normal modes calculated with a coarse-grained elastic network model. Coarse-grained normal mode analysis takes at least an order of magnitude shorter time. Encouraged by this good overlaps and short computation times, we analyzed further low frequency normal modes of Hepatitis C NS5B. Motion directions and average spatial fluctuations have been analyzed in detail. Finally, biological implications of these motions in drug design efforts against Hepatitis C infections have been elaborated.
Only a Subset of Normal Modes Is Sufficient To Identify Linear Correlations in Proteins
(American Chemical Society, 2018) Tekpinar, M.; Yildirim, A.
Identification of correlated residues in proteins is very important for many areas of protein research such as drug design, protein domain classification, signal transmission, allostery and mutational studies. Pairwise residue correlations in proteins can be obtained from experimental and theoretical ensembles. Since it is difficult to obtain proteins in various conformational states experimentally, theoretical methods such as all-atom molecular dynamics simulations and normal-mode analysis are commonly used methods to obtain protein ensembles and, therefore, pairwise residue correlations. The extent of agreement for the correlations obtained with all-atom molecular dynamics and elastic network model based normal-mode analysis is an important issue to investigate due to orders of magnitude computational advantage in terms of wall time for normal-mode based calculation. We performed multiple microsecond long equilibrium classical molecular dynamics simulations for six proteins. We calculated normalized dynamical cross-correlations and linear mutual information as pairwise residue correlations from the trajectories of these simulations. Then, we calculated the same pairwise residue correlations with two elastic network model based normal-mode analysis methods and compared our results with the former. The results show that elastic network model based normal-mode analysis can provide a fast and accurate estimation of linear correlations within proteins. Finally, we observed that only a subset of modes is sufficient to obtain linear correlations in proteins. This conclusion has crucial implications for understanding correlations within very large protein assemblies such as viral capsids. Copyright © 2018 American Chemical Society.