A stochastic, spatially-explicit, demo-genetic model simulating the spread and evolution of a plant pathogen in a heterogeneous landscape to assess resistance deployment strategies.
Loup Rimbaud loup.rimbaud@inrae.fr
Julien Papaix julien.papaix@inrae.fr
Jean-Francois Rey jean-francois.rey@inrae.fr
Jean-Loup Gaussen jean-loup-thomas.gaussen@inrae.fr
Maintainer: Jean-Francois Rey jean-francois.rey@inrae.fr
| Package: | landsepi |
| Type: | Package |
| Version: | 1.1.2 |
| Date: | 2022-03-01 |
| License: | GPL (>=2) |
The landsepi package implements a spatially explicit stochastic model able to assess the epidemiological, evolutionary and economic outcomes of strategies to deploy plant resistance to pathogens. It is based on a spatial geometry for describing the landscape and allocation of different cultivars, a dispersal kernel for the dissemination of the pathogen, and a SEIR (‘susceptible-exposed-infectious-removed’, renamed HLIR for 'healthy-latent-infectious-removed' to avoid confusions with 'susceptible host') structure with a discrete time step. It simulates the spread and evolution of a pathogen in a heterogeneous cropping landscape, across cropping seasons split by host harvests which impose potential bottlenecks to the pathogen.
The lansdcape is represented by a set of polygons where the pathogen can disperse (the basic spatial unit is an individual polygon; an agricultural field may be composed of a single or several polygons). landsepi includes built-in simulated landscapes (and associated dispersal matrices for rust pathogens, see below), but is it possible to use your own landscape (in shapefile format) and dispersal matrix.
A wide array of resistance deployment strategies can be simulated in landsepi: fields of the
landscape are cultivated with different croptypes that can rotate through time; each croptype is
composed of either a pure cultivar or a mixture; and each cultivar may carry one or several resistance
genes. Thus, all combinations of rotations, mosaics, mixtures and pyramiding strategies are
possible. Resistance genes affect several possible pathogen aggressiveness components:
infection rate, durations of the latent period and the infectious period, and propagule
production rate. Resistance may be complete (i.e. complete inhibition of the targeted aggressiveness component) or partial
(i.e. the targeted aggressiveness component is only softened), and expressed from the beginning of the season, or later
(to simulate Adult Plant Resistance (APR), also called Mature Plant Resistance).
Cultivar allocation can be realised via an algorithm (allocateCroptypeCultivars())
but it is possible to use your own cultivar allocation if it is included in the shapefile
containing the landsape.
To each resistance gene in the host (whether it may be a major gene or a QTL for quantitative resistance) is associated a pathogenicity gene in the pathogen. Through mutation of pathogenicity genes, the pathogen can restore its aggressiveness on resistance hosts and thus adapt to resistance (leading to sudden breakdown or gradual erosion of resistance genes). Pathogenicity genes may also be reassorted via sexual reproduction or gene recombination. Increased in aggressiveness on a resistant host (i.e. adaptation to the corresponding resistance genes) can be penalised by a fitness cost on susceptible hosts, i.e. pathogen genotypes adapted to a resistance gene have a reduced aggressiveness on hosts that do not carry this gene. The relation between pathogen aggressiveness on susceptible and resistant hosts is defined by a trade-off relationship whose shape depends on the strength of the trade-off. Strong trade-off means that the gain in fitness on resistant hosts is smaller than the cost on susceptible hosts.
This model provides a useful tool to assess the performance of a wide range of deployment options via epidemiological, evolutionary and economic outputs. It also helps investigate the effect of landscape organisation, the considered pathosystem and the epidemio-evolutionary context on the performance of a given strategy.
The package includes five examples of landscape structures and a default parameterisation to represent
plant pathogens as typified by rusts of cereal crops (genus Puccinia,
e.g. stripe rust, stem rust and leaf rust of wheat and barley).
The main function of the package is runSimul().
It can be parameterised to simulate various resistance deployment strategies using either the provided
landscapes and parameters for cereal rusts, or landscapes and parameters set by the user.
See demo_landsepi() for a demonstration, and our tutorials (browseVignettes("landsepi"))
for details on how to use landsepi.
The spatial unit is a polygon, i.e. a piece of land delimited by boundaries and possibly cultivated with a crop. Such crop may be host or non-host, and the polygon is considered a homogeneous mixture of host individuals (i.e. there is no intra-polygon structuration). A field may be composed of a single or several polygons..
Host individuals are in one of these four categories: H (healthy), E (latent, i.e. infected but not infectious nor symptomatic), I (infectious and symptomatic), or R (removed, i.e. epidemiologically inactive).
A host `individual' is an infection unit and may correspond to a given amount of plant tissue (where a local infection may develop, e.g. fungal lesion) or a whole plant (e.g. systemic viral infection). In the first case, plant growth increases the amount of available plant tissue (hence the number of individuals) during the cropping season. Plant growth is deterministic (logistic growth) and only healthy hosts (state H) contribute to plant growth (castrating pathogen).
The decreasing availability of healthy host tissues (as epidemics spread) makes pathogen infection less likely (i.e. density-dependence due to plant architecture).
Host are cultivated, thus there is no host reproduction, dispersal and natural death.
Environmental and climate conditions are constant, and host individuals of a given genotype are equally susceptible to disease from the first to the last day of every cropping season.
Crop yield depends on the average amount of producing host individuals during the cropping season and does not depend on the time of epidemic peak. Only healthy individuals (state H) contribute to crop yield.
Components of a mixture are independent each other (i.e. there is neither plant-plant interaction nor competition for space, and harvests are segregated).
The pathogen is haploid.
Initially, the pathogen is not adapted to any source of resistance, and is only present on susceptible hosts (at state I).
Pathogen dispersal is isotropic (i.e. equally probable in every direction).
Pathogen reproduction is clonal.
Pathogenicity genes mutate independently from each other.
Pathogen adaptation to a given resistance gene consists in restoring the same aggressiveness component as the one targeted by the resistance gene.
If a fitness cost penalises pathogen adaptation to a given resistance gene, this cost is paid on hosts that do not carry this gene, and consists in a reduction in the same aggressiveness component as the one targeted by the resistance gene.
When there is a delay for activation of a given resistance gene (APR), the time to activation is the same for all hosts carrying this gene and located in the same field.
Variances of the durations of the latent and the infectious periods of the pathogen are not affected by plant resistance.
If there is sexual reproduction (or gene recombination), it occurs only between parental infections located in the same polygon and the same host genotype. The host population is panmictic (i.e. all pairs of parents have the same probability to occur). The propagule production rate of a couple is the sum of the propagule production rates of the parents. The genotype of each daughter propagule is issued from random loci segregation between parental loci.
The epidemiological outcome of a deployment strategy is evaluated using:
the area under the disease progress curve (AUDPC) to measure disease severity (i.e. the average number of diseased plant tissue -status I and R- per time step and square meter),
the relative area under the disease progress curve (AUDPCr) to measure the average proportion of diseased tissue (status I and R) relative to the total number of existing host individuals (H+L+I+R).
the Green Leaf Area (GLA) to measure the average amount of healthy plant tissue (status H) per time step and square meter,
the relative Green Leaf Area (GLAr) to measure the average proportion of healthy tissue (status H) relative to the total number of existing host individuals (H+L+I+R).
the yearly contribution of pathogen genotypes to LIR dynamics on every host as well as the whole landscape.
A set of graphics and a video showing epidemic dynamics can also be generated.
The evolutionary outcome is assessed by measuring:
the dynamics of pathogen genotype frequencies,
the evolution of pathogen aggressiveness,
the durability of resistance genes. Durability can be estimated using the time until the pathogen reaches the three steps to adapt to plant resistance: (1) first appearance of adapted mutants, (2) initial migration to resistant hosts and infection, and (3) broader establishment in the resistant host population (i.e. the point at which extinction becomes unlikely).
The economic outcome of a simulation can be evaluated using:
the crop yield: yearly crop production (e.g. grains, fruits, wine) in weight (or volume) units per hectare (depends on the number of productive hosts and associated theoretical yield),
the crop products: yearly products generated from sales, in monetary units per hectare (depends on crop yield and market value),
the crop operational costs: yearly costs associated with crop planting, in monetary units per hectare (depends on initial host density and planting cost),
the margin, i.e. products - operational costs, in monetary units per hectare.
Future versions:
Future versions of the package will include in particular:
Sets of pathogen parameters to simulate other pathosystems (e.g. canola blackleg, grapevine downy mildew, potato virus Y on pepper).
More flexible initial conditions (e.g. size, location and composition of pathogen inoculum at the beginning of the simulation).
Dependencies:
The package for compiling needs:
g++
libgsl2
libgsl-dev
and the following R packages:
Rcpp
sp
stats
Matrix
mvtnorm
maptools
fields
splancs
sf
DBI
RSQLite
foreach
parallel
doParallel
In addition, to generate videos the package will need ffmpeg.
Rimbaud L., Papaïx J., Rey J.-F., Barrett L. G. and Thrall P. H. (2018). Assessing the durability and efficiency of landscape-based strategies to deploy plant resistance to pathogens. PLoS Computational Biology 14(4):e1006067.
Rimbaud L., Papaïx J. and Rey J.-F. (2018). landsepi: Landscape Epidemiology and Evolution. R package, url: https://cran.r-project.org/package=landsepi.
if (FALSE) {
library("landsepi")
## Run demonstrations (in 20-year simulations) for different deployment strategies:
demo_landsepi(strat = "MO") ## for a mosaic of cultivars
demo_landsepi(strat = "MI") ## for a mixture of cultivars
demo_landsepi(strat = "RO") ## for a rotation of cultivars
demo_landsepi(strat = "PY") ## for a pyramid of resistance genes
}
Run the code above in your browser using DataLab