Nemo

What it simulates

Nemo is a forward-time, individual-based simulator for population genetics and life-history evolution. Its own description calls it genetically explicit: every individual carries its alleles, and patches of individuals can exchange migrants. You compose a simulation from a configuration file that lists a population, a life cycle (the ordered events each generation) and the traits to track.

The example is a quantitative-genetics run. One patch holds 200 hermaphrodites with Wright-Fisher mating. Two traits are controlled by 100 loci, and every locus affects both traits (full pleiotropy). Mutations add normally distributed effects to existing allele values, with rate 0.005 per locus and a mutation matrix of 0.10 variance and 0.02 covariance. Gaussian stabilizing selection pulls both traits toward an optimum of 0 with selection variance 10. A smaller selection variance makes a narrower fitness peak.

Input is one .ini file. Output is a table of statistics at the generations you choose.

When to use it

Use it for questions about how genetic variance and trait covariance build up under mutation, selection and drift, how G-matrices evolve, or how dispersal among patches shapes differentiation. Nemo also covers neutral markers, deleterious mutations and incompatibilities.

Alternatives:

• simuPOP fits better when you want to script the model in Python and control each step.
• SLiM fits better when you want to write the model yourself as a SLiM (Eidos) script and upload it.

Install it yourself

The Nemo site gives conda create -n nemo -c conda-forge -c ecoevo nemo. The recipe uses the bioconda build pinned to 2.4.2, which the anaconda.org index lists for linux-64, linux-aarch64 and osx-arm64.

conda create -n nemo -c conda-forge -c bioconda nemo=2.4.2

A micromamba dry run did not finish (downloads from conda.anaconda.org timed out), so the version rests on the package index listing.

The executable name carries the version. Run a configuration with:

nemo2.4.2 my_simulation.ini

With no argument, Nemo looks for Nemo2.ini in the current directory. The site notes that 2.4.0 and 2.4.1 had a random-number bug affecting free recombination for some bitstring trait architectures, and recommends 2.4.2.

Reading the output

The recipe's run-nemo is a short shell script. It refuses configuration files that set preexec*, postexec* or root_dir (those keys can run shell commands or move the output), then calls nemo2.4.2 on the file.

The run wrote four files:

• quanti.txt (455 bytes) holds the statistics. Columns are replicate, generation, adlt.q1 and adlt.q2 (mean phenotype of each trait among adults), adlt.q1.Va and adlt.q2.Va (within-patch additive genetic variance) and adlt.q12.cov (genetic covariance between the traits).
• quanti.log echoes every parameter, with a start and finish time.
• nemo.log has one summary line per simulation: 300 generations, 1 replicate, Nemo 2.4.2.
• nemo_log.txt is the terminal output, mostly the progress counter.

stat_log_time 100 gives rows at generations 1, 100, 200 and 300. Additive variance of trait 1 grows from 0.085 to 4.30, 5.91 and 8.63. For trait 2 it is 0.103, 5.19, 5.22 and 7.00. The mean phenotypes stay near 0 (between -0.07 and 0.04 for trait 1), which is where the optimum sits. The covariance is 0.017, 0.98, 0.18 and 0.74. Variance is still rising at generation 300, so 300 generations is too short to read an equilibrium.

The file sets random_seed 13215, so the run repeats exactly. A second run gave the same quanti.txt values.

Your own runs

Each run changes one line of the example file.

• selection_variance {{1}} (narrower fitness peak): trait 1 variance at generation 300 is 5.08 instead of 8.63, and trait 2 is 4.39 instead of 7.00.
• quanti_mutation_rate 0.0005 (ten times lower): trait 1 variance at generation 300 is 0.776 instead of 8.63, and trait 2 is 1.15 instead of 7.00.
• replicates 5: replicate 1 matches the single run. At generation 300 the trait 1 variance is 8.63, 7.08, 7.44, 8.33 and 8.21. The file has 20 data rows.
• patch_nbfem 1000 (five times more individuals): trait 1 variance at generation 300 is 9.97 and trait 2 is 10.05.

Pitfalls

• Matrix values need double braces, as in selection_variance {{10}}. The manual says a single decimal applies to all traits and patches, and a matrix gives per-patch and per-trait values.
• The example needs mating_system 6 with patch_nbmal 0. The manual's example for hermaphrodites uses the same pair.
• Rows appear only at stat_log_time multiples. For a curve of Va over time, lower it to 10 or 1.
• One replicate is one draw. In the five-replicate run the trait 1 variance at generation 300 ranged from 7.08 to 8.63.
• The recipe rejects root_dir. In a native run, root_dir sets the directory the results go to.

Further reading

• Nemo home page: https://nemo2.sourceforge.io
• Nemo 2.4.2 manual (PDF): https://sourceforge.net/projects/nemo2/files/Nemo/2.4.2/Nemo2-manual.pdf/download
• Guillaume F, Rougemont J (2006). Bioinformatics 22:2556-2557. https://doi.org/10.1093/bioinformatics/btl415