| 1 |
mmeineke |
95 |
\documentclass[11pt]{article} |
| 2 |
|
|
|
| 3 |
|
|
\usepackage{graphicx} |
| 4 |
|
|
\usepackage{amsmath} |
| 5 |
|
|
\usepackage{amssymb} |
| 6 |
|
|
\usepackage[ref]{overcite} |
| 7 |
|
|
|
| 8 |
|
|
|
| 9 |
|
|
|
| 10 |
|
|
\pagestyle{plain} |
| 11 |
|
|
\pagenumbering{arabic} |
| 12 |
|
|
\oddsidemargin 0.0cm \evensidemargin 0.0cm |
| 13 |
|
|
\topmargin -21pt \headsep 10pt |
| 14 |
|
|
\textheight 9.0in \textwidth 6.5in |
| 15 |
|
|
\brokenpenalty=10000 |
| 16 |
|
|
\renewcommand{\baselinestretch}{1.2} |
| 17 |
|
|
\renewcommand\citemid{\ } % no comma in optional reference note |
| 18 |
|
|
|
| 19 |
|
|
|
| 20 |
|
|
\begin{document} |
| 21 |
|
|
|
| 22 |
|
|
\title{A Mesoscale Model for Phospholipid Simulations} |
| 23 |
|
|
|
| 24 |
|
|
\author{Matthew A. Meineke\\ |
| 25 |
|
|
Department of Chemistry and Biochemistry\\ |
| 26 |
|
|
University of Notre Dame\\ |
| 27 |
|
|
Notre Dame, Indiana 46556} |
| 28 |
|
|
|
| 29 |
|
|
\date{\today} |
| 30 |
|
|
\maketitle |
| 31 |
|
|
|
| 32 |
|
|
\section{Background and Research Goals} |
| 33 |
|
|
|
| 34 |
|
|
\section{Methodology} |
| 35 |
|
|
|
| 36 |
mmeineke |
96 |
\subsection{Length and Time Scale Simplifications} |
| 37 |
mmeineke |
95 |
|
| 38 |
|
|
The length scale simplifications are aimed at increaseing the number |
| 39 |
|
|
of molecules simulated without drastically increasing the |
| 40 |
|
|
computational cost of the system. This is done by a combination of |
| 41 |
|
|
substituting less expensive interactions for expensive ones and |
| 42 |
|
|
decreasing the number of interaction sites per molecule. Namely, |
| 43 |
|
|
charge distributions are replaced with dipoles, and unified atoms are |
| 44 |
|
|
used in place of water and phospholipid head groups. |
| 45 |
|
|
|
| 46 |
|
|
The replacement of charge distributions with dipoles allows us to |
| 47 |
|
|
replace an interaction that has a relatively long range, $\frac{1}{r}$ |
| 48 |
|
|
for the charge charge potential, with that of a relitively short |
| 49 |
|
|
range, $\frac{1}{r^{3}}$ for dipole - dipole potentials |
| 50 |
|
|
(Equations~\ref{eq:dipolePot} and \ref{eq:chargePot}). This allows us |
| 51 |
|
|
to use computaional simplifications algorithms such as Verlet neighbor |
| 52 |
|
|
lists,\cite{allen87:csl} which gives computaional scaling by $N$. This |
| 53 |
|
|
is in comparison to the Ewald sum\cite{leach01:mm} needed to compute |
| 54 |
|
|
the charge - charge interactions which scales at best by $N |
| 55 |
|
|
\ln N$. |
| 56 |
|
|
|
| 57 |
|
|
\begin{equation} |
| 58 |
|
|
V^{\text{dp}}_{ij}(\mathbf{r}_{ij},\boldsymbol{\Omega}_{i}, |
| 59 |
|
|
\boldsymbol{\Omega}_{j}) = \frac{1}{4\pi\epsilon_{0}} \biggl[ |
| 60 |
|
|
\frac{\boldsymbol{\mu}_{i} \cdot \boldsymbol{\mu}_{j}}{r^{3}_{ij}} |
| 61 |
|
|
- |
| 62 |
|
|
\frac{3(\boldsymbol{\mu} \cdot \mathbf{r}_{ij}) % |
| 63 |
|
|
(\boldsymbol{\mu} \cdot \mathbf{r}_{ij}) }{r^{5}_{ij}} \biggr] |
| 64 |
|
|
\label{eq:dipolePot} |
| 65 |
|
|
\end{equation} |
| 66 |
|
|
|
| 67 |
|
|
\begin{equation} |
| 68 |
|
|
V^{\text{ch}}_{ij}(\mathbf{r}_{ij}) = \frac{q_{i}q_{j}}% |
| 69 |
|
|
{4\pi\epsilon_{0} r_{ij}} |
| 70 |
|
|
\label{eq:chargePot} |
| 71 |
|
|
\end{equation} |
| 72 |
|
|
|
| 73 |
|
|
The second step taken to simplify the number of calculationsis to |
| 74 |
|
|
incorporate unified models for groups of atoms. In the case of water, |
| 75 |
|
|
we use the soft sticky dipole (SSD) model developed by |
| 76 |
|
|
Ichiye\cite{Liu96} (Section~\ref{sec:ssdModel}). For the phospholipids, a |
| 77 |
|
|
unified head atom with a dipole will replace the atoms in the head |
| 78 |
|
|
group, while unified $\text{CH}_2$ and $\text{CH}_3$ atoms will |
| 79 |
|
|
replace the alkanes in the tails (Section~\ref{sec:lipidModel}). |
| 80 |
|
|
|
| 81 |
mmeineke |
96 |
The time scale simplifications are taken so that the simulation can |
| 82 |
|
|
take long time steps. By incresing the time steps taken by the |
| 83 |
|
|
simulation, we are able to integrate the simulation trajectory with |
| 84 |
|
|
fewer calculations. However, care must be taken to conserve the energy |
| 85 |
|
|
of the simulation. This is a constraint placed upon the system by |
| 86 |
|
|
simulating in the microcanonical ensemble. In practice, this means |
| 87 |
|
|
taking steps small enough to resolve all motion in the system without |
| 88 |
|
|
accidently moving an object too far along a repulsive energy surface |
| 89 |
|
|
before it feels the affect of the surface. |
| 90 |
mmeineke |
95 |
|
| 91 |
mmeineke |
96 |
In our simulation we have chosen to constrain all bonds to be of fixed |
| 92 |
|
|
length. This means the bonds are no longer allowed to vibrate about |
| 93 |
|
|
their equilibrium positions, typically the fastest periodical motion |
| 94 |
|
|
in a dynamics simulation. By taking this action, we are able to take |
| 95 |
|
|
time steps of 3 fs while still maintaining constant energy. This is in |
| 96 |
|
|
contrast to the 1 fs time step typically needed to conserve energy when |
| 97 |
|
|
bonds are allowed to vibrate. |
| 98 |
mmeineke |
95 |
|
| 99 |
|
|
\subsection{The Soft Sticky Water Model} |
| 100 |
|
|
\label{sec:ssdModel} |
| 101 |
|
|
|
| 102 |
mmeineke |
96 |
|
| 103 |
|
|
|
| 104 |
|
|
\begin{equation} |
| 105 |
|
|
\label{eq:ssdTotPot} |
| 106 |
|
|
V_{\text{ssd}} = V_{\text{LJ}}(r_{i\!j}) + V_{\text{dp}}(\mathbf{r}_{i\!j}, |
| 107 |
|
|
\boldsymbol{\Omega}_{i},\boldsymbol{\Omega}_{j}) |
| 108 |
|
|
+ V_{\text{sp}}(\mathbf{r}_{i\!j},\boldsymbol{\Omega}_{i}, |
| 109 |
|
|
\boldsymbol{\Omega}_{j}) |
| 110 |
|
|
\end{equation} |
| 111 |
|
|
|
| 112 |
|
|
\begin{equation} |
| 113 |
|
|
\label{eq:spPot} |
| 114 |
|
|
V_{\text{sp}}(\mathbf{r}_{i\!j},\boldsymbol{\Omega}_{i}, |
| 115 |
|
|
\boldsymbol{\Omega}_{j}) = |
| 116 |
|
|
v^{\circ}[s(r_{ij})w_{ij}(\mathbf{r}_{ij},\boldsymbol{\Omega}_{i}, |
| 117 |
|
|
\boldsymbol{\Omega}_{j}) |
| 118 |
|
|
+ |
| 119 |
|
|
s'(r_{ij})w^{x}_{ij}(\mathbf{r}_{ij},\boldsymbol{\Omega}_{i}, |
| 120 |
|
|
\boldsymbol{\Omega}_{j})] |
| 121 |
|
|
\end{equation} |
| 122 |
|
|
|
| 123 |
|
|
\begin{equation} |
| 124 |
|
|
\label{eq:apPot2} |
| 125 |
|
|
w_{ij}(\mathbf{r}_{ij},\boldsymbol{\Omega}_{i},\boldsymbol{\Omega}_{j}) = |
| 126 |
|
|
\sin\theta_{ij} \sin 2\theta_{ij} \cos 2\phi_{ij} |
| 127 |
|
|
+ \sin \theta_{ji} \sin 2\theta_{ji} \cos 2\phi_{ji} |
| 128 |
|
|
\end{equation} |
| 129 |
|
|
|
| 130 |
|
|
\begin{equation} |
| 131 |
|
|
\label{eq:spCorrection} |
| 132 |
|
|
\begin{split} |
| 133 |
|
|
w^{x}_{ij}(\mathbf{r}_{ij},\boldsymbol{\Omega}_{i},\boldsymbol{\Omega}_{j}) &= |
| 134 |
|
|
(\cos\theta_{ij}-0.6)^2(\cos\theta_{ij} + 0.8)^2 \\ |
| 135 |
|
|
&\phantom{=} + (\cos\theta_{ji}-0.6)^2(\cos\theta_{ji} + 0.8)^2 - 2w^{\circ} |
| 136 |
|
|
\end{split} |
| 137 |
|
|
\end{equation} |
| 138 |
|
|
|
| 139 |
|
|
\begin{equation} |
| 140 |
|
|
\label{eq:spCutoff} |
| 141 |
|
|
s(r_{ij}) = |
| 142 |
|
|
\begin{cases} |
| 143 |
|
|
1& \text{if $r_{ij} < r_{L}$}, \\ |
| 144 |
|
|
\frac{(r_{U} - r_{ij})^2 (r_{U} + 2r_{ij} |
| 145 |
|
|
- 3r_{L})}{(r_{U}-r_{L})^3}& |
| 146 |
|
|
\text{if $r_{L} \leq r_{ij} \leq r_{U}$},\\ |
| 147 |
|
|
0& \text{if $r_{ij} \geq r_{U}$}. |
| 148 |
|
|
\end{cases} |
| 149 |
|
|
\end{equation} |
| 150 |
|
|
|
| 151 |
mmeineke |
95 |
\subsection{The Phospholipid Model} |
| 152 |
|
|
\label{sec:lipidModel} |
| 153 |
|
|
|
| 154 |
|
|
|
| 155 |
|
|
\bibliographystyle{achemso} |
| 156 |
|
|
\bibliography{canidacy_paper} |
| 157 |
|
|
\end{document} |
