Difference between revisions of "Publications"

J. Stat. Mech. 084010, 2019.

One of my longest-running projects is finally published. Large deviation functions calculated from "experimental" simulations of two model systems illustrate the key ideas of maximum transition entropy - a refinement of Jaynes' maximum caliber that preserves causality.

We extend four forcefields to model hydroxypropylcellulose and find that none of them can reproduce experimentally measured hydration forces from 2001. Long-range solvent-mediated forces remain challenging targets for predictive simulation because of a breakdown in science funding priorities.

This work presents parallel histories of the development of two modern theories of condensed matter: the theory of electron structure in quantum mechanics, and the theory of liquid structure in statistical mechanics. Key developments provide some guidance on important directions for future advancements in theory and practice.

I re-write Ornstein-Zernike theory in a convenient matrix basis so that future computational implementations can be made robust and that numerical and fitting error can be tightly controlled. It is hoped that the framework can be used in the near future to compute molecular solvation free energies and to realize its enormous potential economic and environmental benefits for formulating industrial fluids and consumer products.

My co-authors apply my algorithm and software implementation of Bayesian generalized linear model regression with linear inequality constraints to estimate molecular modeling parameters from quantum their mechanical calculations. They find the method gives robust results that greatly increase efficiency of additional simulations.

We present a way to utilize Essmann et. al.'s smooth particle mesh Ewald algorithm (implemented here) to get the low-angle direct correlation function from canonical ensemble simulation data. The result avoids well-documented issues with long wavelength (small-angle scattering) modes that appear when using real-space methods. This result is combined with new estimates of finite-size effects and grand-canonical ensemble corrections to provide definitive results on the problem of extrapolating Kirkwood-Buff integrals.

We compare neutron reflectivity experiments to molecular dynamics calculations on the orientation and binding position of the Dengue envelope protein responsible for viral escape from the host cell's endosome. In addition to the hydrophobic fusion tip, we find important roles for three positively-charged residues in the viral protein that contribute to host membrane binding. These could potentially be targeted by new anti-viral medicines.

This paper presents a new, simple ansatz for adding dissipation to arbitrary stochastic forcing of a quantum dynamical system. For Gaussian random forces, it predicts a Lindblad equation identical to the Caldeira-Leggett model up to order <math>\beta^2</math>, where the same term is present with a different prefactor. The system-centric, phase space picture here shows that the <math>\beta^2</math> term represents a quantum confinement effect.

This paper provides measurement-based definitions of heat and work that can be realized in current laboratory setups. The first and second laws are proved despite the fact that temperature is treated completely as as a property of the interacting reservoir. Measurements of the work are subject to the famous EPR paradox because the work exchanged between two quantum systems is not defined until a measurement is performed. Based on this, we show that even an environment at absolute zero cannot lower a system's temperature below a minimum characteristic of the way the environment is coupled to the system.

We set out to test the Joule-Thomson analysis of thermodynamics of porous flow for gasses through a nanopore and found that while local equilibrium is established in the steady-state, finite-size effects cause heat flow opposite the flow direction that violates the assumption of an adiabatic porous plug.

This paper introduces 3 basic functions that generalize BLAS to tensors and presents a code generation strategy for their efficient execution on GPUs that achieves peak performance on the same order of magnitude as for traditional, vendor-optimized matrix-multiplications.

An elegant solution to the problem of finding periodic images in non-rectangular lattices is provided based on the closest vector problem. Related code is here: [1]

This paper gives multiple experimental measurements of binding energy between the Dengue virus envelope protein and host membranes that largely confirm our computational predictions from 2015. I contributed all the theory for terminal velocity during sedimentation, along with a novel kinetic analysis providing the free energy and enthalpy of the dissociation barrier (all the details are at the end of the appendix).

The work presents a new, general, theoretical foundation for the dynamics of open quantum systems modeled on the maximum entropy derivation of equilibrium statistical mechanics. Computational results are presented for three detailed systems to validate and reinforce the theory. It represents a significant advancement for the field, as it lucidly connects the dynamics of a single wavefunction plus environmental noise to the Caldeira-Leggett model for density matrices.

J. Mol. Biol. 427:2867-2885, 2015.

We present an explanation for the (until now controversial) catalytic mechanism of type 2 bacterial L-asparaginase enzymes. By using the carboxylic acid of the substrate (asparagine) as the proton acceptor, this enzyme is able to preferentially carry out deamidation on asparagine more quickly than for the competing substrate, glutamine. The hypothesis, re-discovered from our MD simulations, was first put forward years ago in contested experimental studies, and now finds additional support from our MD and QM calculations.

arXiv:1509.07036 submitted, 2015.

This paper describes the implementation of a new interpreted language for distributed parallel computing. It achieves its goal by maintaining pure functional semantics, allowing all terms in the language to be partially evaluated and serialized to network storage at any point during computation.

We studied the influence of dispersion energy corrections on the free energy of formation for sodium-water clusters computed with DFT and wound up discovering that dispersion and split-range exchange functionals can somewhat counter-balance each other. The charged sodium ion pulls on the water's electrons, clearly showing which density functionals over-polarize compared to CCSD. Split-range exchange can reduce this over-polarization, but results in reduced electrostatic interaction. Dispersion can lower the binding energy again to counter-balance. So, functionals fit to experimental formation energies need both effects to avoid lowering energies by over-polarizing.

I introduce sets of point charges that are able to simultaneously reproduce all multipole (spherical harmonic) expansions up to arbitrary order. The number of points is space-optimal. Translations are described from the usual harmonics and from Cartesian moments (dipole, quadrupole, etc.) on supersymmetric tensors to directional moments using the point weight distribution. Many applications are possible, including trivial implementation of multipoles in molecular mechanics and representing probability distributions over rotation space.

A fully atomistic potential of mean force for association of the viral envelope protein from Dengue virus was compared to a Poisson-Boltzmann electrostatic plus dispersion model. The results are consistent, showing hope for this type of combined scale simulation.

This paper with our experimental collaborators carried out blocked peptide synthesis by vapor-depositing Boc-L-alanine to create a uniform thin film of polypeptides grown on a silica substrate activated by aminopropyltrimethoxysilane.

We show how the interpretation of thermodynamic states as representing system information leads naturally to thermodynamic cycles and the first and second laws of thermodynamics as well as similar formulations for nontrivial nonequilibrium problems. The logical development of the theory also leads naturally to correct indistinguishability factors in the partition function.

We review the development of models for understanding the physical basis of selectivity for K+ ions over Na+, its sibling only one row behind, in membrane channels and transporters. Although the problem is subtle because of the morass of competing effects, we emphasize work analyzing the systematic influence of the environment on tipping local binding site structure toward selective configurations.

We presented an extension of the Quasi-Chemical theory for quantifying the impact of local structure on ion complexation thermodynamics. The theory can be simply represented using a set of thermodynamic cycles involving binding site structural and compositional states as reaction intermediates.

We summarize work on designing polymer coatings for salt exclusion in water transporting nanopores. In this work, I collected available molecular dynamics results for these systems and performed a novel energy efficiency analysis able to relate atomistic and experimental scales as well as identify important design goals and chemical principles for material performance.

We investigated the central quantity of free energies in a Bayesian context and provide estimators for solvation free energies as well as optimal potential of mean force approximations to model polymer coarse-grained dynamics from atomistic simulations.

Dr. Zhao's ab-initio analysis of the charge distribution in water-ion clusters highlighted the importance of many-body water-water interactions and charge transfer effects in determining cluster structural and energetic properties. These are still challenging to represent in modern polarizable forcefields and have implications for anion properties at interfaces.

The role of polarizability in forcefield-based models of ions and water was examined. Utilizing some of our recent developments on quasi-chemical theory, we have been able to quantify the tightened, asymmetric nature of the ion's local solvation waters induced by increased polarizability as well as the exact effects of polarization on the solvation free energy. The results suggest some potential problems and diagnostics for such models.

This force matching software implements and tests coarse-graining for general molecular systems in a mere 4000 lines of code. It is able to parametrize coarse Hamiltonians from atomic trajectory data given arbitrary definitions of coarse united-atom type models as well as carry out short Langevin Dynamics simulations on the coarse scale. The program's main drawbacks are its slow speed and high memory usage due to its simplistic design, attributable to the interpreted nature of python.

Fresh insight is provided into long-standing mathematical issues surrounding computational modeling of continuous functions from a few sampled data points. The present research lays the groundwork for predicting the behavior of complicated many-body systems using advanced regression techniques.

We develop a Bayesian method for computing (with error bars) the free energy for forming a nano-bubble in an arbitrary solvent system. This forms the first step of a thermodynamic cycle for dissolving a real solute. We prove that upper and lower bounds for that solvation free energy can be obtained from two simulations (with and without the solute present). The method is excellent for dissolving gas in water, while the upper/lower bounds are larger for dissolving water or ions.