plotFitnessLandscape: Plot a fitness landscape.
In OncoSimulR: Forward Genetic Simulation of Cancer Progression with Epistasis

Description Usage Arguments Value Note Author(s) References See Also Examples

View source: R/fitness_landscape_utils.R

Show a plot of a fitness landscape. The plot is modeled after (actually, mostly a blatant copy of) that of MAGELLAN, http://wwwabi.snv.jussieu.fr/public/Magellan/.

Note: this is not a plot of the fitnessEffects object; for that, see plot.fitnessEffects.

plotFitnessLandscape(x, show_labels = TRUE,
                     col = c("green4", "red", "yellow"),
                     lty = c(1, 2, 3),
                     use_ggrepel = FALSE,
                     log = FALSE, max_num_genotypes = 2000,
                     only_accessible = FALSE,
                     accessible_th = 0,
                     ...)

## S3 method for class 'genotype_fitness_matrix'
plot(x, show_labels = TRUE,
                                   col = c("green4", "red", "yellow"),
                                   lty = c(1, 2, 3),
                                   use_ggrepel = FALSE,
                                   log = FALSE, max_num_genotypes = 2000,
                                   only_accessible = FALSE,
                                   accessible_th = 0,
                                   ...)

## S3 method for class 'evalAllGenotypes'
plot(x, show_labels = TRUE,
                                   col = c("green4", "red", "yellow"),
                                   lty = c(1, 2, 3),
                                   use_ggrepel = FALSE,
                                   log = FALSE, max_num_genotypes = 2000,
                                   only_accessible = FALSE,
                                   accessible_th = 0,
                                   ...)

## S3 method for class 'evalAllGenotypesMut'
plot(x, show_labels = TRUE,
                                   col = c("green4", "red", "yellow"),
                                   lty = c(1, 2, 3),
                                   use_ggrepel = FALSE,
                                   log = FALSE, max_num_genotypes = 2000,
                                   only_accessible = FALSE,
                                   accessible_th = 0,
                                   ...)

`x`	One of the following: A matrix (or data frame) with g + 1 columns. Each of the first g columns contains a 1 or a 0 indicating that the gene of that column is mutated or not. Column g+ 1 contains the fitness values. This is, for instance, the output you will get from `rfitness`. A two column data frame. The second column is fitness, and the first column are genotypes, given as a character vector. For instance, a row "A, B" would mean the genotype with both A and B mutated. The output from a call to `evalAllGenotypes`. Make sure you use `order = FALSE` in that call. The output from a call to `evalAllGenotypesMut`. Make sure you use `order = FALSE`. The output from a call to `allFitnessEffects`. The first two are the same as the format for the `genotFitness` component in `allFitnessEffects`.
`show_labels`	If TRUE, show the genotype labels.
`col`	A three-element vector that gives the colors to use for increase, decreases and no changes in fitness, respectively. The first two colours are also used for peaks and sinks.
`lty`	A three-element vector that gives the line types to use for increase, decreases and no changes in fitness, respectively.
`use_ggrepel`	If TRUE, use the ggrepel package to avoid overlap of labels.
`log`	Log-scale the y axis.
`max_num_genotypes`	Maximum allowed number of genotypes. For some types of input, we make a call to `evalAllGenotypes`, and use this as the maximum.
`only_accessible`	If TRUE, show only accessible paths. A path is considered accesible if, at each mutational step (i.e., with the addition of each mutation) fitness increases by at least `accessible_th`. If you set `only_accessible = TRUE`, the number of genotypes displayed can be much smaller than the number of existing genotypes if many of those genotypes are not accessible via any path.
`accessible_th`	The threshold for the minimal change in fitness at each mutation step (i.e., between successive genotypes) to be used if `only_accessible = TRUE`.
`...`	Other arguments passed to `plot`. Not used for now.

A fitness landscape plot: a plot showing paths between genotypes and peaks and sinks (local maxima and minima).

I have copied most of the ideas (and colors, and labels) of this plot from MAGELLAN (http://wwwabi.snv.jussieu.fr/public/Magellan/) but MAGELLAN has other functionality that is not provided here such as epistasis stats for the landscape, and several visual manipulation options.

One feature of this function that is not available in MAGELLAN is showing genotype labels (i.e., annotated by gene names), which can be helpful if the different genotypes mean something to you.

In addition to the above differences, another difference between this plot and those of MAGELLAN is how sinks/peaks of more than one genotype are dealt with. This plot will show as sinks or peaks sets of one or more genotypes that are of identical fitness (and separated by a Haming distance of one). So a sink or a peak might actually be made of more than one genotype. In MAGELLAN, as far as I can tell, peaks and sinks are always made of a single isolated genotype.

Does this matter? In most realistic cases where not two genotypes can have exactly the same fittnes it does not. In some cases, though, it might matter. Are multi-genotype sinks/peaks really sinks/peaks? Arguably yes: suppose genotypes "AB" and "ABC" both have fitness 0, which is minimal among the fitness in the set of genotypes, and genotypes "A" and "ABCD" have fitness 0.1. To go from "A" to "ABCD", if you want to travel through "AB", you have to go through the valley of "AB" and "ABC"; once in "ABC" you can climb up to "ABCD"; and once in "AB" you can move to "ABC" since it has identical fitness to "AB". Mutatis mutandis for multi-genotype peaks. Ignoring the possibility of peaks/sinks made of more than one genotype actually makes code much simpler.

Sometimes not showing the any links that involve a decrease in fitness can help see non-accessible pathways (in strong selection, no multiple mutations, etc); do this by passing, for instance, an NA for the second element of col.

Finally, use common sense: for instance, if you pass a allFitnessEffects that specifies for, say, the fitness of a total of 5000 genotypes you'll have to wait a while for the plot to finish.

Ramon Diaz-Uriarte

MAGELLAN web site: http://wwwabi.snv.jussieu.fr/public/Magellan/

Brouillet, S. et al. (2015). MAGELLAN: a tool to explore small fitness landscapes. bioRxiv, 31583. http://doi.org/10.1101/031583

allFitnessEffects, evalAllGenotypes, allFitnessEffects, rfitness, plot.fitnessEffects

## Generate random fitness for four genes-genotypes
## and plot landscape.

r1 <- rfitness(4)
plot(r1)


## Specify fitness in a matrix, and plot it

m5 <- cbind(A = c(0, 1, 0, 1), B = c(0, 0, 1, 1), F = c(1, 2, 3, 5.5))
plotFitnessLandscape(m5)


## Specify fitness with allFitnessEffects, and plot it

fe <- allFitnessEffects(epistasis = c("a : b" = 0.3,
                                          "b : c" = 0.5),
                            noIntGenes = c("e" = 0.1))

plot(evalAllGenotypes(fe, order = FALSE))

## same as
plotFitnessLandscape(evalAllGenotypes(fe, order = FALSE))