An architecture maps a genome to genotypic values. Genomes are matrices of allelic effect vectors (traits x alleles); alleles \(2l - 1\) and \(2l\) belong to locus \(l\), and the effect of a locus is the sum of its two alleles (\(Y_l\)). Effects are measured from the reference genotype, where all effects are zero.
Arguments
- loci
number of diploid loci.
- traits
number of traits.
- init_b
probability that an entry of
Bis 1 in a new population (0.25 in the C code, which starts from random mutations of an emptyB).- init_y
how allelic effects
yof a new population are drawn, with the model'sinit_sd:"normal",Normal(0, init_sd)for every allele;"mutations", as in the C code,2 x locimutations ofNormal(0, init_sd)added to alleles chosen at random in each genome.- genes
number of weight genes.
- init_weights
starting values of the weight genes (one per gene).
- init_contributions
starting contribution matrix (
traits x genes); by default ones on the diagonal (genejaffects traitjonly, the identity whengenes == traits).- epsilon
epistasis coefficients: a
traits x traits x traitsarray shared by all pairs of loci (symmetric in its last two indices; seerandom_epsilon()), or atraits x traits x traits x loci x lociarray with a different tensor for every pair of loci, as in Jones, Burger & Arnold (2014) (seepairwise_epsilon()). For the pairwise form,epsilon[k, a, b, i, j]multipliesY_i[a] Y_j[b]in traitk, and only entries withi < jare used.
Details
additive(): \(x = \sum_l Y_l\).binary_b(): the binary ontogenetic matrix model of Melo & Marroig (2015). Each allele has a scalar effect \(y\) and a column of a binary matrix \(B\) saying which traits it affects; \(x = B y\). Use withbinary_b_mutation(). Genomes havetraits + 1rows: \(y\), then \(B\).contribution_matrix(): the model of do O & Whitlock (2023).genesweight genes \(a_j\) and atraits x genesmatrix of contribution genes \(c_{ij}\), each its own locus with a scalar value (the two alleles of a locus are summed); \(x_i = \sum_j c_{ij} a_j\). Pleiotropy (\(C\)) evolves continuously and can be negative, and the map is bilinear (epistasis between weight and contribution genes). Loci1..genesare the weight genes; locusgenes + (j - 1) * traits + iis \(c_{ij}\). Use withcontribution_mutation(). New populations start with every individual atinit_weightsandinit_contributions(each allele half of the locus value), plusNormal(0, init_sd)noise on every allele.multilinear()with a pairwiseepsilon(pairwise_epsilon()): a different tensor for every pair of loci (Jones, Burger & Arnold 2014), \(x = S + \sum_{i<j} \varepsilon_{ij}(Y_i, Y_j)\).multilinear(): pairwise multilinear epistasis (Hansen & Wagner 2001), \(x = S + \sum_{i<j} \varepsilon(Y_i, Y_j)\), with \(S = \sum_l Y_l\) and \(\varepsilon(a, b)_k = \sum_{l,m} \varepsilon_{klm} a_l b_m\). The same tensor is used for every pair of loci and must be symmetric in its last two indices.
Examples
additive(loci = 50, traits = 2)
#> <evomod architecture> additive - 50 loci, 2 traits
binary_b(loci = 50, traits = 4)
#> <evomod architecture> binary_b - 50 loci, 4 traits
contribution_matrix(traits = 4)
#> <evomod architecture> contribution_matrix - 20 loci, 4 traits
multilinear(loci = 50, traits = 2, epsilon = random_epsilon(2, sd = 0.05))
#> <evomod architecture> multilinear - 50 loci, 2 traits