A Trojan horse approach to medical intervention strategies
Do you have questions or comments about this model? Ask them here! (You'll first need to log in.)
A Trojan horse approach to medical intervention strategies
A [NetLogo] model by Rik Blok.
http://www.zoology.ubc.ca/~rikblok/wiki/doku.php?id=science:popmod:brown2009:start
This agent-based model represents a bacterial infection as described in [[Brown2009]]. It is assumed the wild type (W) produces a public good (F) which benefits all. To combat the infection the population is innoculated with a "cheater" strain (C). Changes are simply represented as elementary reactions between (local) agents:
- N → 2 N @ rate 1 (growth)
- W + N → N @ rate 1 (competition)
C + N → N @ rate 1 - W → W + F @ rate b (public good)
W + F → 2 W @ rate β - W → ∅ @ rate x (wild cost)
- C → ∅ @ rate q (cheater cost)
where N indicates an individual of either type and ∅ indicates an absence of products.
Additionally, the cheater strain has one or more of the following traits:
- It consumes but does not produce the public good (F),
C + F → 2 C @ rate β - It produces a toxin (T) that harms all,
C → C + T @ rate a
N + T → ∅ @ rate δ - It produces a bacteriocinogen (B) it is immune to but harms the wild type,
C → C + B @ rate e
C + B → C @ rate γ
W + B → ∅ @ rate γ.
The target and infected cells are fixed whereas the virions move randomly with diffusion constant diff-const. If 'well-mixed' is on then they are shuffled randomly in space with each timestep.
Finally, it is assumed that the wild type is more "virulent" (harmful to the host) than the cheater strain.
How it works
The simulation approximates a Poisson process for each of the above events. The best known technique would be the Gillespie algorithm [[Gibson2000]] but it isn't well suited to NetLogo's strengths. Instead, time proceeds in steps with multiple events occurring in each timestep.
The step size is adaptive, chosen to achieve a desired error tolerance, compared with the Gillespie algorithm. When the error tolerance is near zero the likelihood of each event is small and we may expect just a few events to occur per timestep. Then we're accurately -- but inefficiently -- mimicking the Gillespie algorithm. As the tolerance increases we have more simultaneous events, lowering accuracy but increasing performance.
References
[[Brown2009]] Brown, Sam P., West, Stuart A., Diggle, Stephen P., and Griffin, Ashleigh S. Social evolution in micro-organisms and a Trojan horse approach to medical intervention strategies. Philosophical Transactions of the Royal Society B: Biological Sciences, 364: 3157-3168. doi:10.1098/rstb.2009.0055. 2009.
[[Gibson2000]] Gibson, Michael A. and Bruck, Jehoshua. Efficient Exact Stochastic Simulation of Chemical Systems with Many Species and Many Channels. J. Phys. Chem. A, 104(9): 1876-1889. doi:10.1021/jp993732q. 2000.
[[NetLogo]] Wilensky, U. NetLogo. http://ccl.northwestern.edu/netlogo/. Center for Connected Learning and Computer-Based Modeling, Northwestern University. Evanston, IL. 1999.
Comments and Questions
; Revisions: ; 2013-08-02 ; - accuracy now reports ( 100% - per-capita-mean-error ) ; - diffuse-individual moved into .nls file ; - output-wrap moved into .nls file ; - sigfigs moved into .nls file ; 2013-07-29 ; - line-breaking procedure to replace output-prints ; - moved some generic code into includes file ; - events are defined at beginning and applied behind the scenes (in included file) ; design choices: ; * the (full) model is agent-based ; * bacteria (wilds and cheaters) are mobile (Brownian) ; * their products (food, toxin, and bacteriocide) are stationary ; * interactions are local (per patch) ; * time steps forward in increments large enough for multiple events to occur ; * the full model fixes the fast (food, toxin, & bacteriocide) production rates ; * the System Dynamics Modeler contains a numeric approximation ; * the numeric model applies the QSSA to reduce the system, removing the products ; * the numeric model approximates the equations in Brown et al., 2009 __includes [ "diffuse-individual.nls" "output-wrap.nls" "per-capita-tau-leaping.nls" "sigfigs.nls" ] ; the possible "breeds" of turtles breed [wilds wild] breed [cheaters cheater] ; turtles own event rates patches-own [ food toxin bacteriocide ] globals [ count-patches ; assign to variable to improve speed; used in setup and plots beta-food delta-toxin gamma-bacteriocide fast-rate ; = beta = delta = gamma (fast consumption rates for food, toxin, & bacteriocide) wild-virulence ] to startup setup end to setup clear-all set fast-rate 5 ; TO DO: how big does this need to be to satisfy QSSA? set wild-virulence 2 ; assume wild type twice as virulent as cheaters (as per Brown et al.) set beta-food fast-rate set delta-toxin fast-rate set gamma-bacteriocide fast-rate set count-patches count patches system-dynamics-setup ; setup system-dynamics-modeler for numeric comparison ; set breed shapes set-default-shape turtles "circle" ; create numbers of each breed to get the desired average density (per patch) create-wilds ( W-num * count-patches ) [ set color green ] create-cheaters ( C-num * count-patches ) [ set color red ] ; shuffle them ask turtles [ setxy random-xcor random-ycor ] ask patches [ ; start with QSSA densities of food, toxin, & bacteriocide let count-turtles-here count turtles-here let count-cheaters-here count cheaters-here if ( count-turtles-here > 0 ) [ ; Poisson-distributed random variates set food random-poisson ( b-shared-benefit * count wilds-here / ( beta-food * count-turtles-here ) ) set toxin random-poisson ( a-toxin-production * count-cheaters-here / ( delta-toxin * count-turtles-here ) ) set bacteriocide random-poisson ( e-bacteriocinogen * count-cheaters-here / ( gamma-bacteriocide * count-turtles-here ) ) ] ] reset-ticks ; hint output-wrap 26 ; set wrapping column output-wrap "An agent-based model of 'Social evolution in micro-organisms and a Trojan horse approach to medical intervention strategies' by Brown et al. (2009)." output-wrap "Press go to get started.\n" ; merge all events into one master list [2013-07-29] add-per-capita-event turtles ; N -> 2 N @ rate 1 task [ 1 ] task [ hatch 1 ] add-per-capita-event turtles ; N + N -> N @ rate 1 task [ count other turtles-here ] task [ die ] add-per-capita-event turtles ; N + F -> 2 N @ rate beta task [ beta-food * food ] task [ set food food - 1 hatch 1 ] add-per-capita-event turtles ; N + T -> 0 @ rate delta task [ delta-toxin * toxin ] task [ set toxin toxin - 1 die ] add-per-capita-event wilds ; W -> 0 @ rate x task [ x-wild-cost ] task [ die ] add-per-capita-event wilds ; W -> W + F @ rate b task [ b-shared-benefit ] task [ set food food + 1 ] add-per-capita-event wilds ; W + B -> 0 @ rate gamma task [ gamma-bacteriocide * bacteriocide ] task [ set bacteriocide bacteriocide - 1 die ] add-per-capita-event cheaters ; C -> 0 @ rate q task [ q-cheater-cost ] task [ die ] add-per-capita-event cheaters ; C -> C + T @ rate a task [ a-toxin-production ] task [ set toxin toxin + 1 ] add-per-capita-event cheaters ; C -> C + B @ rate e task [ e-bacteriocinogen ] task [ set bacteriocide bacteriocide + 1 ] add-per-capita-event cheaters ; C + B -> C @ rate gamma task [ gamma-bacteriocide * bacteriocide ] task [ set bacteriocide bacteriocide - 1 ] end to go if ticks = 0 [ reset-timer output-wrap "You may adjust all parameters while the simulation runs." output-wrap "Compare the numeric, non-spatial (faint) vs. agent-based (bold) dynamics in the graphs." output-wrap ( word "It is assumed that the wild type is " wild-virulence " times as virulent as the cheater.\n" ) ] system-dynamics-go ; numeric model set dt per-capita-tau-leap inaccuracy dt diffuse-individual turtles diffusion-const dt do-plots if (dt = 0) [ ; simulation over output-wrap (word "Finished in " timer " seconds.") stop ] end to do-plots set-current-plot "dynamics" system-dynamics-do-plot update-plots ; update other plots ("Edit" the plots to see code) end to-report diffusion-const report ifelse-value well-mixed [ 0 ][ 10 ^ log-diffusion ] end to-report inaccuracy ; complement of accuracy, depends on err-tolerance report 10 ^ ( err-tolerance ) end
There are 4 versions of this model.
Attached files
File | Type | Description | Last updated | |
---|---|---|---|---|
A Trojan horse approach to medical intervention strategies.png | preview | Preview for 'A Trojan horse approach to medical intervention strategies' | over 11 years ago, by Rik Blok | Download |
diffuse-individual.nls | extension | Random walk for individuals. | over 11 years ago, by Rik Blok | Download |
output-wrap.nls | extension | Same as output-print but wraps printed output. | over 11 years ago, by Rik Blok | Download |
per-capita-tau-leaping.nls | extension | A variant tau-leaping algorithm to approximate ; Gillespie's Stochastic Simulation Algorithm (SSA). | over 11 years ago, by Rik Blok | Download |
sigfigs.nls | extension | Like precision but specifies number of significant figures. | over 11 years ago, by Rik Blok | Download |
This model does not have any ancestors.
This model does not have any descendants.