Smoldyn in the research literature

All papers that mention Smoldyn (Google scholar)

Published research that used Smoldyn simulations (internal link)

Publications that describe Smoldyn software and algorithms (internal link)

Published research that used Smoldyn simulations

neural synapse

Computational investigation of the changing patterns of subtype specific NMDA receptor activation during physiological glutametergic neurotransmission

Singh, P., A.J. Hockenberry, V. Tiruvadi, and D.F. Meaney, PLoS Comput. Biol. 7:e1002106, 2011

Neural NMDA receptors mediate many physiological functions, including the molecular basis for learning and memory. These receptors exist in various subtypes, which enables them to discriminate between different types of signals. Using Smoldyn models, the authors found that different receptors have different dynamic ranges, that specific subtypes dominate in long-term depression and long-term potentiation situations, and that the content of a specific subtype enhances response magnitude and fidelity during long-term potentiation. The yellow and red portions of the figure show pre- and post-synaptic regions, while dots represent receptors.
neural spine

Sequestration of CaMKII in dendritic spines in silico

Khan, S., Y. Zou, A. Amjad, A. Gardezi, C.L. Smith, C. Winters, T.S. Reese, J. Comput. Neurosci. In press, 2011

This work investigated the sequestration of calcium dependent kinase II (CamKII) in neural dendritic spines following synaptic stimulation. It looked at the effects of spine geometry on CamKII diffusion, binding of CamKII to the post-synaptic density, to the cytoskeleton, and to other CamKII proteins. Simulation results compared favorably with microscopy experiments. This work showed that self-aggregation of CamKII could provide a switch that amplifies CamKII sequestration and regulates its activity.
Ultrasensitivity

Ultrasensitivity in Multisite Phosphorylation of Membrane-Anchored Proteins

Dushek, O., P.A. van der Merwe, and V. Shahrezaei, Biophys. J. 100:1189-1197, 2011

The authors showed that multiple phosphorylation sites on a single protein can produce switch-like responses when reactions between proteins are limited by diffusion. This is the case for many reactions that occur on the plasma membrane. The figure shows that ultrasensitivity increases with as the number of protein phosphorylation sites increase.
Bar1 model

Detailed Simulations of Cell Biology with Smoldyn 2.1

Andrews, S.S., N.J. Addy, R. Brent, and A.P. Arkin, PLoS Comp. Biol. 6:e1000705, 2010

The authors investigated mating pheromone signaling between yeast cells. A central "receiver" cell is covered with receptors that bind mating pheromone, which is secreted by surrounding "sender" cells (in the figure, unbound receptors are blue, bound receptors are red, and pheromone is green). This work showed that the central cell can better locate strong pheromone emitting cells when the central cell secretes a pheromone-degrading protease because the protease cloud sharpens the local pheromone gradient.
Simulated FRAP

Introducing simulated cellular architecture to the quantitative analysis of fluorescent microscopy

DePristo, M.A., L. Chang, R.D. Vale, S.M. Khan, and K. Lipkow, Prog. Biophys. Mol. Biol. 100:25-32, 2009

The authors used simulations to analyze and quantify experimental Fluorescence Recovery After Photobleaching (FRAP) data for proteins in the E. coli chemotaxis system. They quantified protein diffusion coefficients of 2 μm2/s and assessed turnover rates between cytoplasmic proteins and membrane-associated protein clusters. The figure shows simulated CheY-YFP fluorescence in a protein cluster as it is photobleached and then recovers from protein exchange. In general, simulations can be useful tools for the quantitative analysis of fluorescent microscopy experiments.
Protein gradients

Model for Protein Concentration Gradients in the Cytoplasm

Lipkow, K. and D.J. Odde, Cell. Mol. Bioeng. 1:84-92, 2008

Using simulations, the authors showed that several of the E. coli chemotaxis proteins (CheY and CheZ) likely have stable protein concentration gradients across the length of cells. This arises from protein activation that is localized to one cell pole and activation-dependent protein complexation that affects protein diffusion coefficients. These mechanisms, and similar ones, occur for many other proteins as well, so stable protein concentration gradients within cells are likely to be widespread.
Hair cell PMCA2

Rapid Turnover of Stereocilia Membrane Proteins: Evidence from the Trafficking and Mobility of Plasma Membrane Ca2+-ATPase 2

Grati, M., M.E. Schneider, K. Lipkow, E.E. Strehler, R.J. Wenthold, and B. Kachar, J. Neuroscience 26:6386-6395, 2006

This work combined experimental and modeling approaches to investigate membrane protein turnover in stereocilia (fine hairs in the inner ear that convert mechanical motion to electrical signals) to better understand stereocilia recovery from trauma, such as loud noises. The authors focused on the spatial distribution, mobility, and trafficking of the PMCA2 protein, which is both abundant and essential to stereocilia function. They found that PMCA2 exhibits rapid turnover, which supports other evidence that stereocilia undergo rapid continuous renewal. The figure shows PMCA2 (green dots) distribution on an unrolled stereocilia membrane over time.
Lotka Volterra scale Lotka Volterra dynamics

Simulating cell biology

Andrews, S.S. and A.P. Arkin, Current Biology 16:R523-R527, 2006

This is a primer on how to simulate cell biology systems. It presents the basics of ODE (ordinary differential equation) modeling, stochastic modeling, spatial modeling, and spatial-stochastic modeling. Of particular interest, it presents simulations of the same Lotka-Volterra predator-prey system that used each of the different methods. The results differ dramatically. Deterministic simulations show nearly sinusoidal oscillations, non-spatial stochastic simulations show oscillations with increasing amplitudes, and particle-tracking (Smoldyn; figure show here) simulations show a characteristic boom-and-bust pattern.
CheZ localization

Changing Cellular Location of CheZ Predicted by Molecular Simulations

Lipkow, K., PLoS Comp. Biol. 2:e39, 2006

In the E. coli chemotaxis system, the CheY messenger protein is phosphorylated at a receptor cluster near one cell pole and is dephosphorylated by the CheZ protein. The author used simulations to explore the signaling behaviors that arise with different CheZ localizations. She found that the model that best agreed with experiments included CheZ proteins that oligomerize and localize to the receptor cluster upon cellular stimulation. The black dots in the figure are CheZ oligomers (also, pink dots are CheY, red are CheYp, and green are non-oligomeric CheZ). This creates a negative feedback loop which improves system precision, robustness, and adaptation.
FliM occupancy

Simulated Diffusion of Phosphorylated CheY through the Cytoplasm of Escherichia coli

Lipkow, K., S.S. Andrews, and D. Bray, J. Bact. 187:45-53, 2005

The authors developed a spatial stochastic simulation of the central E. coli chemotaxis system. It includes the core proteins and reactions, a polar-localized receptor cluster, and several flagellar motors in the cell membrane. Increasing cytoplasmic crowding, due to ribosomes and other macromolecules, resulted in slower protein diffusion, steepened concentration gradients, and substantial differences in signal propagation delays to motors at different positions (figure).

Publications that describe Smoldyn software and algorithms

GPU Smoldyn

Smoldyn on Graphics Processing Units: Massively Parallel Brownian Dynamics Simulation

Dematté, L., IEEE Trans. Comput. Biol. Bioinform. in press, 2011

The author parallelized the core Smoldyn code to run on Graphics Processing Units (GPUs). This can accelerate simulations by factors of up to 130, thus enabling the efficient simulation of models that would be impractical to run on conventional computer hardware.
Smoldyn example

Spatial and stochastic cellular modeling with the Smoldyn simulator

Andrews, S.S., Methods in Molecular Biology in press, 2011

This paper describes the use of the Smoldyn simulation program. It focuses on the basic concepts, while leaving syntax details to the Smoldyn User's Manual. This figure was taken from an example simulation that presents the use of reactions, molecule-surface interactions, compartments, and other Smoldyn features. This paper presents an especially useful Notes section that gives suggestions on choosing simulation parameters (e.g. diffusion coefficient and reaction rates) and methods for improving computational performance.
Smoldyn accuracy

Detailed Simulations of Cell Biology with Smoldyn 2.1

Andrews, S.S., N.J. Addy, R. Brent, and A.P. Arkin, PLoS Comp. Biol. 6:e1000705, 2010

This paper (which is also listed above) describes the Smoldyn software. It gives an overview of the software algorithms, features, and performance. It shows that Smoldyn is very accurate, is more than twice as fast as either MCell or ChemCell, and has a wide range of features. This figure shows the number of molecules adsorbed to a surface over time for several adsorption coefficients; close accuracy between simulation data (dots) and theoretical calculations (lines) show Smoldyn's accuracy. The inset shows that simulation noise matches theoretical calculations.
Adsorption

Accurate particle-based simulation of adsorption, desorption and partial transmission

Andrews, S.S., Phys. Biol. 6:046015, 2009

This paper describes methods for quantitatively simulating the adsorption of molecules to surfaces, the desorption of molecules from surfaces, and the partial transmission of molecules through surfaces. Central results of the paper include look-up tables that give a molecule's adsorption or transmission probability based on the surface's adsorption or transmission coefficient, the molecule's diffusion coefficient, the simulation time step, and the rate of the reverse desorption or transmission process. Smoldyn uses these algorithms for molecule-surface interactions.
Simulator comparison

Computational methods for diffusion-influenced biochemical reactions

Dobrzyński, M., J.V. Rodríguez, J.A. Kaandorp, and J.G. Blom, Bioinformatics 23:1969-1977, 2007

The authors compared the three major spatial stochastic simulation methods, which are particle-based methods (with Smoldyn), spatial Gillespie methods (with GMP), and Green's Function Reaction Dynamics (GFRD; with custom code). They found that GFRD is most accurate and that the relative accuracy between particle-based and spatial Gillepsie methods depends on the latter's spatial resolution; with fine resolution, spatial Gillespie methods can be better than particle-based methods. For systems with many molecules, GFRD is extremely slow, while Smoldyn takes about twice as long as GMP to yield results with comparable accuracy. The authors noted that particle-based methods are particularly flexible for creating realistic geometries.
Bimolecular reaction simulation

Stochastic simulation of chemical reactions with spatial resolution and single molecule detail

Andrews, S.S. and D. Bray, Phys. Biol. 1:137, 2004

The authors derived several of the central algorithms for particle-based simulation, all designed for high accuracy and good computational efficiency. Methods are presented for simulating diffusion and chemical reactions, where the reactions can be 0th, 1st, or 2nd order and they can be irreversible or reversible. This figure shows one of the paper's central results, which is a look-up table that enables simulated bimolecular chemical reaction rates (y-axis) to be determined from simulation parameters (x-axis) for both irreversible (bottom line) and reversible reactions (other lines).