nobrien97/genetic-architecture-in-slim

Tracking evolution in forward-modelling software SLiM

The effects of genetic constraint and linked selection on polygenic adaptation

This branch uses SLiM 3.6 (https://messerlab.org/slim/) to model how the mutational constraint on genes can influence polygenic adaptation to a new fitness optimum under a quantitative additive model for a single trait.

Each gene has a specified level of constraint, which determines the ratios of mutation types that can occur at each site. This approximates the classic figures from neutral theory we have all seen:

Courtesy of Blackwell Publishing¹

Where each gene has an individual one of these figures, and they can be of differing shape depending on constraint. For example, highly constrained genes (such as those encoding histones) will extremely rarely mutate favourably, so their mutations are mostly deleterious (and greatly so), with fewer neutral mutations, and almost no beneficial mutations. Trait fitness is calculated based on additive phenotype effects from trait mutations based on Lande's (1976)² Gaussian fitness model. These phenotype effects are sampled from a normal distribution, N(0, σ), where σ is a parameter varied across treatments.

Populations are challenged to adapt to a new phenotypic optimum from a burnt-in starting position. A variety of genetic parameters are adjusted including

• Population size
• Proportion of the genome under high, medium, and low constraint
• Genome-wide recombination rate
• The number of QTLs contributing to the trait
• The strength of selection
• The size of the phenotypic shift (initial distance to the optimum)
• Variance in additive phenotype effects

To explore this parameter space, I use Latin hypercube sampling, courtesy of the R packages DoE.wrapper and LHS. I then run each parameter combination 48 times with seeds generated from my own Mersenne Twister based 64-bit seed generator, seedgenerator, which can be found in ./src/Tools/SeedGenerator/.

References

¹ Ridley, 2003, Evolution, Blackwell Publishing, https://www.blackwellpublishing.com/ridley/images/neutral_theory.jpg

² Lande, 1976, Natural Selection and Random Genetic Drift in Phenotypic Evolution, Evolution 30:314-334

genetic-architecture-in-slim's Issues

Mutation output incomplete in one line only

This seems pretty rare, has only happened once: mutation output: didn't put sim.generation at the front of one line in mutation outputs for the SFS_Test HPC run.
The total file is 191093 rows, and row 9835 lacks a sim.generation value.

The line references a mutation object which only just appeared ( 108 generations prior). The same mutation in the entry just prior (gen 78200 rather than 78300; line 9581) has no such issue, so it's unlikely to be a rogue if statement causing problems. Perhaps it's a problem with the write itself?