Genespace Simulation Model Description
Genespace is a spatially and genetically explicit individual-based simulation. Generations are discrete and non-overlapping. Events occur in the following order: birth, mating, dispersal, reproduction, death. The simulation is written in C++ and uses pseudorandom numbers generated using the mt19937 Mersenne Twister and an algorithm by Vose (1991). In the sections below, user-specified parameters are capitalized and given in italics. A full parameter list is given here. Further details and source code are available on request.
Individuals
- Individuals are diploid with separate sexes and have a number of genetic traits. Mate choice based on multiple criteria is simulated, with each criterion influenced by up to three traits. Each male display is represented as a single additive trait, m. Two other traits, termed f, and c, also influence mate choice for each mating criterion. Trait f specifies the display phenotype of males preferred by females, while trait c determines the strength and direction of mating preference (mating discrimination).
- An ecological trait termed x influences survival. Two ecological optima, THETA_S and THETA_E, can be specified for this trait; if the optima are different, ecotype formation or speciation occurs in some conditions.
- Each trait is determined by L additive independent diallelic loci with equal effect size. All traits are scaled between 0 and 1. Mutation rate is MUT_RATE for trait loci. In addition, 8 neutral microsatellite loci are simulated with stepwise mutation rate MMUT_RATE to assess the barrier to gene flow between morphs.
Spatial organization
- The population is simulated as a two-dimensional array of demes of size DEMESxDEMES. Demes are simulated with toroid (TOROID = 1) or closed (TOROID = 0) boundaries. Individuals disperse to an adjacent deme with MIG_RATE % probability. Demes are uninhabitable with probability DEAD_SITES %.
- Uninhabitable demes are reorganized randomly each iteration with probability S_I.
- Parameter ENVSCE controls the initial scenario. Scenarios are
- ENVSCE = 0. A hybrid zone, with two populations initially, each adapted to its own ecological optimum x=THETA_S or x=THETA_E.
- ENVSCE =1. A single population initially adapted to THETA_S .
- Habitat patches: If patch size I_H is > 0, two separate habitats will be simulated and selection will only favor THETA_S in one habitat and THETA_E in the other habitat. These habitats are distributed in patches of size S_H. If patch size I_H = 0 , only one habitat exists and selection favors THETA_S and THETA_E simultaneously.
Selection
- Survival of individuals is regulated using the Beverton-Holt model, which is influenced by individual fitness, the total number of local juveniles N, the mean number of offspring of each individual B, and deme carrying capacity K.
Reproduction
- Females choose mates by randomly sampling males in the same deme until one was accepted. When choice is based on multiple criteria, a male must be accepted based on all criteria. Females who have not chosen a mate after sampling MMM males mate with the last male sampled. Costs of choosiness are represented by a probability of not mating COST which is multiplied by the mean deviation of the female's c traits from 0.5. This implies that more discriminating females are less likely to breed, but does not select against females who prefer rare male traits.
- Mated females produce a Poisson-distributed number of offspring with mean B.
- The simulation uses a mating probability function first proposed by Lande (1981) and further developed by Gavrilets (2004, p.321), and Gavrilets (2007).
- For each mating criterion, the probability of mating is Gaussian and influenced by the three phenotypic traits, m (display trait of male), f (ecological trait preferred by female) and c (mating discrimination trait).
- Parameter SIGMA_A scales the strength of mating preferences by altering the variance of the distribution of mating probabilities.
- More extreme c phenotypes create stronger preferences. Females with c=0.5 mate randomly. Trait f determines the value of the male trait (m) preferred by a female. Trait c generates a preference for x similar to f when c>0.5, and for x dissimilar to f when c<0.5.
- Non-random mating systems are specified as follows.
- Matching Traits m, f and c are all independent traits. This means that females can evolve to prefer males with large or small m, regardless of their own m. For example, a combination of either small f and c<0.5, or large f and c>0.5 trait values would generate positive assortative mating in individuals with large m, but negative assortative mating in individuals with small m. Conversely, either large f with c<0.5, or small f with c>0.5 generates positive assortative mating with small m but negative assortative mating with large m.
- Open Traits m and c are independent traits, while f has a fixed value of one. Females with c>0.5 prefer mates who have the maximum possible m, whereas females with c<0.5 prefer mates who have the minimum possible m. Functionally, this model closely resembles previous models of open-ended preference (e.g. \citealt{gav04}, p. 321). Under this model, positive assortative mating results when c>0.5 for females with large m, and when c<0.5 for females with small m; negative assortative mating results when c<0.5 for females with large m, and when c>0.5 for females with small m.
The male display trait may be "magic" or "arbitrary", as follows:
- Magic The male display trait m is also the ecological trait, x. In speciation literature, when disruptive ecological selection drives divergence, this scenario is often termed a "magic trait" model because no linkage disequilibrium is needed for speciation.
- Arbitrary Male display trait m is independent of the ecological trait, x. The display maybe neutral or subject to stabilizing selection. This is the scenario used for classic Fisherian runaway selection; its contribution to speciation has long been disputed.
Data reporting
- The simulation results are displayed graphically in real time.
- Data are also recorded as tab-delimited data in a text file called output.txt.
- You can specify how often data are written to the text file by modifying the settings file, settings.txt. Note that outputting to a text file is SLOW, and files can get very large.
- Text output consists of a line of text for each trait in each iteration recorded. The data recorded on each line are:
- Parameter values for all settings
- Time (number of iterations since the simulation started)
- Trait (X, M, F, C)
- Result (classifies phenotype distribution as "unimodal" (only one mode), "bimodal" (two or more modes), or "species" (non-zero at both extremes and zero in the middle, implying no F1 hybrids were present)
- FST. If >SAMPLE individuals were present at both extremes of the phenotype distribution, this value measures the genetic divergence between the two morphs at neutral microsatellite loci; otherwise, it is replaced by an *.
- Frequency value for each possible phenotype of this trait. For example, this would be 9 values in a 4-locus system, or 17 values in an 8-locus system.
- Descriptions of the parameters and their effects in the model are given here