Overview

FREGENE is a C++ program that simulates sequence-like data over large genomic regions in large diploid populations. Unlike coalescent-based simulation tools, such as MS (Hudson, 2002), FREGENE works forwards-in-time which allows a wide range of demographic and selection scenarios to be implemented. Many such models are already incorporated into FREGENE, and since it is open source users can modify or extend these. Coalescent methods have difficulty incorporating large amounts of gene conversion or crossover (Hoggart et al. 2007), whereas these pose no particular problem for FREGENE. FREGENE offers a flexible model for recombination hotspots, and can readily simulate regions up to tens of Mb on a standard desktop computer.

The principle limitation of forward-in-time algorithms is computational, since the entire population must be tracked through time, not only the chromosomes that are ancestral to the observed sample. FREGENE implements many features to enhance computational efficiency, and includes a rescaling option that greatly reduces computation time at the cost of some approximation.

The program SAMPLE, that comes with the FREGENE package, generates samples of individuals from a FREGENE output population, together with a phenotype that depends on the genotype at one or more SNPs and may be binary (case/control) or Gaussian. SAMPLE can also summarize the SNP minor allele frequency (MAF) spectrum and calculate $r^2$ values for SNP pairs.

For further details about FREGENE, including the rescaling, see:

Balding DJ (2008). FREGENE: Simulation of realistic sequence-level data in populations and ascertained samples . BMC Bioinformatics in press.

Hoggart CJ, Chadeau-Hyam M, Clark TG, Lampariello R, Whittaker JC, De Iorio M, Balding DJ (2007) Sequence-level population simulations over large genomic regions. Genetics 177: 1725-1731, 2007, doi: 10.1534/genetics.106.069088

Please cite these articles in any publication that uses FREGENE.

FREGENE is free to use, distribute and modify, under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or any later version. In particular FREGENE comes WITHOUT ANY WARRANTY.

Please report any problems or bugs to d.balding@ic.ac.uk

Summary of FREGENE's modelling assumptions

FREGENE simulates a, possibly subdivided, population of monoecious, diploid, individuals whose genomes consist of a single, linear chromosome. The population evolves over non-overlapping generations according to a Wright-Fisher model, with or without selection.

• Mutation: a two-allele, symmetric mutation model: multiple mutations at a site are allowed; the mutation rate is the same for each allele at each site.
• Recombination: both crossovers and gene conversions, with rates that may be uniform or vary along the chromosome at both broad and fine scales (the latter corresponding to hotspots).
• Population size: constant, or exponential growth.
• Migration: within a subdivided population; the islands can have different population sizes.
• Selection: A pre-specified proportion of sites is non-neutral. At sites under selection, fitnesses are allocated to each genotype stochastically, according to distributions that can allow directional (positive or negative) and balancing selection scenarios. Selection operates additively over sites. In a divided population, it can be local to the subpopulation in which the selected mutant arose, or global.

Many of these assumptions are easy to relax by small changes to the source code, but typically at a cost in computational efficiency.

Features of the implementation

• For computational efficiency, each chromosome is represented as a list of sites at which the minor allele is present
  • If a derived allele becomes the major allele, then the allele is ``swapped'' so that the ancestral allele is now recorded.
  • FREGENE outputs the ``swap'' status of each site, i.e. whether the minor allele is derived or ancestral.
• The main FREGENE output file, in xml format, summarizes the parameters of the simulation and includes the chromosomes present in the final generation. It is formatted so that it can be re-processed as an input file, which allows complex demographic scenarios to be built up from simple components, via repeat calls of FREGENE.
• FREGENE requires about 2 days on a standard desktop computer to simulate the evolution of a 10 Mb chromosome in 10K individuals over 100K generations. However, by rescaling the number of generations, population size, and the rate parameters, the computing time can be greatly reduced at the cost of some approximation. For example, with ten-fold rescaling the above simulation time is reduced by a factor of 64 to less than 1 hour. Disadvantages of rescaling include:
  1. Reduced population size; however, by setting a switch the user can ask FREGENE to run extra generations in which the rescaling is relaxed, bringing the population size up to its desired level.
  2. More mutations arising at sites that are already polymorphic (``double hit'', or ``back'' mutations); FREGENE tracks these so this effect can be monitored.

See Hoggart et al. (2007) for further details.