Sampling Performance of Multiple Independent Molecular Dynamics Simulations of an RNA Aptamer

Abstract

Using multiple independent simulations instead of one long simulation has been shown to improve the sampling performance attained with the molecular dynamics (MD) simulation method. However, it is generally not known how long each independent simulation should be, how many independent simulations should be used, or to what extent either of these factors affects the overall sampling performance achieved for a given system. The goal of the present study was to assess the sampling performance of multiple independent MD simulations, where each independent simulation begins from a different initial molecular conformation. For this purpose, we used an RNA aptamer that is 25 nucleotides long as a case study. The initial conformations of the aptamer are derived from six de novo predicted 3D structures. Each of the six de novo predicted structures is energy minimized in solution and equilibrated with MD simulations at high temperature. Ten conformations from these six high-temperature equilibration runs are selected as initial conformations for further simulations at ambient temperature. In total, we conducted 60 independent MD simulations, each with a duration of 100 ns, to study the conformation and dynamics of the aptamer. For each group of 10 independent simulations that originated from a particular de novo predicted structure, we evaluated the potential energy distribution of the RNA and used recurrence quantification analysis to examine the sampling of RNA conformational transitions. To assess the impact of starting from different de novo predicted structures, we computed the density of structure projection on principal components to compare the regions sampled by the different groups of ten independent simulations. The recurrence rate and dependence of initial conformation among the groups were also compared. We stress the necessity of using different initial configurations as simulation starting points by showing long simulations from different initial structures suffer from being trapped in different states. Finally, we summarized the sampling efficiency for the complete set of 60 independent simulations and determined regions of under-sampling on the potential energy landscape. The results suggest that conducting multiple independent simulations using a diverse set of de novo predicted structures is a promising approach to achieve sufficient sampling. This approach avoids undesirable outcomes, such as the problem of the RNA aptamer being trapped in a local minimum. For others wishing to conduct multiple independent simulations, the analysis protocol presented in this study is a guide for examining overall sampling and determining if more simulations are necessary for sufficient sampling.

Prediction of Aptamer–Small-Molecule Interactions Using Metastable States from Multiple Independent Molecular Dynamics Simulations

“…To carry out a comprehensive molecular docking analysis, diverse representative structures of an unbound aptamer should be considered. Previous works have shown that short independent simulations sample a larger conformational space rather than a unique trajectory. , Furthermore, simulations parting from different initial configurations can explore dynamics and energy basins difficult to sample from a single initial conformation . Therefore, the objective of this study is to assess the binding affinity by molecular docking of metastable representative structures of DNA aptamers obtained from molecular dynamics (MD) simulation of distinct initial conformations.…”

Molecular Dynamics Simulations of the Aptamer Domain of Guanidinium Ion Binding Riboswitch ykkC-III: Structural Insights into the Discrimination of Cognate and Alternate Ligands

“…Both these stages were accomplished using 25,000 steepest descent (SD) steps and 25,000 conjugate gradient (CG) minimization steps. Each system was then heated to 300 K in a 100 ps simulation performed at a constant volume with the RNA restrained using a 10 kcal mol –1 Å –2 force constant, using the Langevin temperature equilibration scheme. , This was followed by five independent 100 ns simulations on each of the four aptamer states using different initial velocities, which amount to a total of 2 μs of production simulation time . All production simulations were carried out in an NPT ensemble maintained at 300 K temperature and 1 atm pressure, using the Langevin thermostat and Berendsen barostat, respectively.…”

“…59,60 This was followed by five independent 100 ns simulations on each of the four aptamer states using different initial velocities, which amount to a total of 2 μs of production simulation time. 61 All production simulations were carried out in an NPT ensemble maintained at 300 K temperature and 1 atm pressure, using the Langevin thermostat and Berendsen barostat, 62 respectively. The bonds involving hydrogen atoms were constrained using the SHAKE algorithm during the heating and production steps.…”

Electrochemical Aptasensors for Antibiotics Detection: Recent Achievements and Applications for Monitoring Food Safety

“…Although docking simulations are powerful techniques to elucidate 3D structures that are stable with respect to energetics, ideally one would also study the dynamics of these structures over time. MD simulations generate a conformational ensemble of nucleic acid structures at equilibrium and provide detailed atomic motions that aid in understanding structure–function relationships in an aptamer [ 112 , 114 ]. Due to the advantages of incorporating MD simulations into a computational pipeline, an increasing number of published studies report the use of software packages such as NAMD, AMBER, and GROMACS [ 101 , 115 , 116 ].…”

“…Due to the advantages of incorporating MD simulations into a computational pipeline, an increasing number of published studies report the use of software packages such as NAMD, AMBER, and GROMACS [ 101 , 115 , 116 ]. Ultimately, computational methods could be applied to predict the interactions between an aptamer and its ligand and, hence, refine the nucleic acid sequence to improve the binding affinity and specificity, in addition to being applied for the de novo selection of aptamer in silico [ 114 , 116 ]. An overview of the workflow followed for the simulation-based prediction of an aptamer’s interactions with, and affinity for, its ligand is shown in Figure 2 .…”