Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 Aug 25;120(33):8276-88.
doi: 10.1021/acs.jpcb.6b01991. Epub 2016 Apr 20.

Rapid Characterization of Allosteric Networks With Ensemble Normal Mode Analysis

Affiliations
Free PMC article

Rapid Characterization of Allosteric Networks With Ensemble Normal Mode Analysis

Xin-Qiu Yao et al. J Phys Chem B. .
Free PMC article

Abstract

Allosteric regulation is a primary means of controlling protein function. By definition, allostery involves the propagation of structural dynamic changes between distal protein sites that yields a functional change. Gaining improved knowledge of these fundamental mechanisms is important for understanding many biomolecular processes and for guiding protein engineering and drug design efforts. In this work we compare and contrast a range of normal mode analysis (NMA) approaches together with network analysis for the prediction of structural dynamics and allosteric sites. Application to heterotrimeric G proteins, hemoglobin, and caspase 7 indicates that atomistic elastic network models provide improved predictions of experimental allosteric mutation sites. Results for G proteins also display an improved consistency with those derived from more computationally demanding MD simulations. Application of this approach across available experimental structures for a given protein family in a unified manner, that we refer to as ensemble NMA, yields the best overall predictive performance. We propose that this atomistic ensemble NMA approach represents an efficient and powerful tool for guiding the exploration of coupled motions and allosteric mechanisms in cases where multiple structures are available and where MD may prove prohibitively expensive.

Figures

Figure 1
Figure 1. Differences in ensemble NMA predicted local and global dynamics distinguish functional states of Gα
(A) Characterization of distinct GTP, GDP, and GDI states from a clustering of NMA derived RMSIP values. (B) Fluctuation analysis reveals structural regions with significantly distinct flexibilities among GTP (red), GDP (green) and GDI (blue) states (sites with a p-value < 0.05 are highlighted with a blue shaded background). For each state, ensemble averaged values are shown. Short vertical tick lines at the top of each plot indicate the location of nucleotide binding site residues. (C) Projection of crystallographic structures in the PC1-PC2 subspace of the PCA performed on correlation matrices (See text for details). (D) Identification of residue couplings that characterize specific functional states. Couplings with significantly high (positive) or low (negative) weights in PC1 of the correlation based PCA are indicated with blue and red colors respectively. These couplings are also shown mapped to molecular structure.
Figure 2
Figure 2. Convergence of structural dynamics in MD simulations
Simulations are performed under three distinct states: GTP (A), GDP (B) and GDI (C). Under each state, five metrics are calculated to compare distinct subsets of simulations: atomic fluctuation, covariance, network community, network node centrality and network suboptimal path. Metric value for MDn (n=1, 2, 3, 4 or 5) is calculated using n independently performed 80-ns simulation trajectories, either simply an ensemble average (fluctuation and covariance) or output of ensemble derived network (community, centrality and path; See Methods). Comparison is made between values for the same metric but obtained from two non-overlapping sets of n simulations.
Figure 3
Figure 3. Ensemble NMA has overall improved performance when compared to single structure based NMA
Difference prediction scores of fluctuation, covariance, community, centrality and path between ensemble NMA and single structure NMA are shown for GTP (A), GDP (B) and GDI (C) bound states. Seven methods are considered in the comparison, including six ENMs (mfAAENM, mfHCA, AAENM, ANM1.8, HCA, and sdENM) and the all-atom NMA (AmberNMA). The predominantly positive bars indicate that the ensemble approach has overall improved performance in relation to single structure based calculations. Comparisons in which both approaches have low prediction scores (<0.6) are excluded. The maximum score across methods for each metric is also shown in parentheses at the bottom of each plot.
Figure 4
Figure 4. Optimization of the number of modes used for correlation network construction in ENMs
SIP scores are computed between the centralities derived from HCA (A) or AAENM (B) networks and those derived from MD networks.
Figure 5
Figure 5. Structures of hemoglobin and caspase 7 with catalytic and allosteric sites shown
(A) Hemoglobin tetramer (constructed based on the PDB structure 2DN1) with each subunit displayed in distinct color. The oxygen molecules (red) and experimentally verified allosteric sites (yellow) are also labeled. (B) Caspase 7 homodimer (PDB code 1F1J) with regulatory (C-terminal of L2’) and Cys-His dyad catalytic sites colored in cyan. Experimentally verified allosteric sites are depicted as yellow sticks.
Figure 6
Figure 6. Ensemble NMA predicts state specific local and global dynamics of hemoglobin
(A) Characterization of distinct “Tense” (T) and “Relaxed” (R) states from a clustering of NMA (HCA) derived RMSIP values. (B) Fluctuation analysis reveals structural regions with significantly distinct flexibilities between T and R states (sites with a p-value < 0.005 are highlighted with a blue shaded background). For each state, ensemble averaged values are shown. Short vertical tick lines at the top of each plot indicate the location of heme binding site residues. Red asterisks at the top indicate the experimentally verified allosteric sites. (C) Projection of crystallographic structures in the PC1–PC2 subspace of the PCA performed on correlation matrices. (D) Identification of residue couplings that characterize specific functional states. Couplings with significantly high (positive) or low (negative) weights in PC1 of the correlation based PCA are indicated with green and red colors respectively. These couplings are also shown mapped to molecular structure.
Figure 7
Figure 7. Ensemble NMA predictions for caspase 7
(A) Characterization of distinct “Active”, “Intermediate”, and “Inactive” states from a clustering of NMA (HCA) derived RMSIP values. (B) Fluctuation analysis reveals structural regions with distinct flexibilities across states. Short vertical tick lines at the top of each plot indicate the location of the loop L2’ (regulatory site) and the catalytic site residues. Red asterisks at the top indicate the experimentally verified allosteric sites. (C) Projection of crystallographic structures in the PC1–PC2 subspace of the PCA performed on correlation matrices. (D) Identification of residue couplings that characterize specific functional states. Couplings with significantly high (positive) or low (negative) weights in PC1 of the correlation based PCA are indicated with green and red colors respectively. These couplings are also shown mapped to the molecular structure.

Similar articles

See all similar articles

Cited by 6 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback