Developing and applying new theoretical and computational methods to study complex condensed phase systems
Material for Download
Multi-scale Coarse-graining (MS-CG) Force Matching (FM) code is now publicly available for download
Immature HIV-1 lattice assembly dynamics are regulated by scaffolding from nucleic acid and the plasma membrane
In this work, we use coarse-grained molecular models to simulate the early stages of viral assembly, packaging, and budding during HIV-1 replication. In collaboration with the Lippincott-Schwartz Group at the NIH and HHMI, we provide a molecular-level view into the dynamics of this process and show that Gag polyproteins engage with viral RNA and deformations along the cell membrane to both instigate and control its self-assembly into the immature HIV-1 lattice. We further demonstrate the importance of specific features of the modeled Gag domains. Taken together, our findings elucidate a simple regulatory network of interactions that may contain viable targets for antivirals.
In this work, we show that using minimal bias methods, we can simulate a protein in solution and have it behave as if embedded in a different environment, such as a large protein complex. This is done by learning a bias on coarse-grained observables, such as distances or angles between beads representing large collections of atoms in the protein. The bias is targeted such that these observables sample the correct mean and variance from simulations in the whole complex. We apply this to the protein actin in solution, and discuss how this would be useful for future expensive sampling or QM/MM simulations performed on actin, previously done with rigid constraints that pollute the protein's dynamics.
In this work, we apply the Experiment Directed Simulation (EDS) method to improve the properties of water simulated with Ab Initio Molecular Dynamics (AIMD) simulations at a poor/cheap DFT level of theory. A simple classical bias is learned on-the-fly from the O-O radial distribution function, which causes the AIMD water to match much better the structural properties of water seen in experiment, including diffusion. Applying this bias to a hydrated excess proton in water also significantly improves the properties of that system, without the need for learning a new bias potential.
Development of Reactive Force Fields Using Ab Initio Molecular Dynamics Simulation Minimally Biased to Experimental Data
By using relative entropy minimization (REM) and minimally biased ab initio molecular dynamics (AIMD) simulations, we have developed two new multiscale reactive molecular dynamics force fields for a hydrated excess proton. We show that both of these models closely reproduce the solvation structure of the reference AIMD data, and we also demonstrate the capabilities of REM to develop reactive force-fields.
Extending the Range and Physical Accuracy of Coarse-grained Models: Order Parameter Dependent Interactions
The choice of basis function is an important decision that must go into the construction of a coarse-grained model. Coarse-grained models are usually built using a basis set of pairwise interactions for non-bonded interactions and expanding into 3-body and on to expand the basis set. This paper introduces the use of order parameters, particularly local density and absolute position, as an efficient alternative to expand basis sets. The order parameter basis set allows for a drastically improved description of liquid-vapor interfaces in coarse-grained methanol.
Internal symmetry in transmembrane protein — exact or approximate — is common and it has consequences for the Gaussian fluctuations around the equilibrium structure of the proteins. We substantiate that a coarse-grained mapping must preserve the underlying structural symmetries (both structural symmetry groups and modular repeats in the secondary structure, depending on the resolution) and show that this can be used as a design principle to construct coarse-grained mappings.
ClC-ec1, a Cl- /H+ antiporter, is critical for maintaining ion concentrations and PH gradients in bacteria in acidic environments. In this work, we computationally characterized the rate-limiting step of the overall proton transport process in ClC-ec1 and the essential mechanism of the Cl-/H+ coupling. We found that the highest barrier for PT is located at the deprotonation of E148, and this barrier is significantly reduced by the binding of Cl- in the central site, which displaces E148 and thereby facilitates its deprotonation.
Molecular Modeling and Assignment of IR Spectra of the Hydrated Excess Proton in Isotopically Dilute Water
We developed a mixed quantum-classical model for the vibrational spectroscopy of the excess proton in isotopically dilute water. The model is useful for decomposing IR spectra into contributions from different aqueous proton configurations as validated by our experimental collaborator Andrei Tokmakoff (UChicago). We find that the shift from Eigen to Zundel-like configurations is distinguished by a decrease in the O—H transition frequency.
The influenza A M2 channel (AM2) transports protons into the influenza virus upon acid activation. MS-RMD simulations were performed to characterize the free energy profiles of the proton transport events in the M2 channel. Our results show that decreasing pH causes the Trp41 gate to open, which decreases the deprotonation barrier of the His37 tetrad. This leads to channel activation, which is characterized by increased proton conductance.
Transition-Tempered Metadynamics is a Promising Tool for Studying the Permeation of Drug-like Molecules through Membranes
The recently developed transition-tempered metadynamics (TTMetaD) has been proven to converge asymptotically without sacrificing exploration of the collective variable space in the early stages of simulations. We applied TTMetaD to study the permeation of drug-like molecules through a lipid bilayer to investigate its usefulness in medicinal chemistry. Compared to other enhanced sampling methods, TTMetaD is able to predict the most accurate and reliable estimate of the potential of mean force in the early stages of the simulations. We also show that using multiple randomly initialized replicas allows convergence analysis and provides an efficient means to converge the simulations in shorter wall times and CPU times.