Computational protocol: Genetic and maternal effects on tail spine and body length in the invasive spiny water flea (Bythotrephes longimanus)

Similar protocols

Protocol publication

[…] Using our clonal breeding design, we quantified genetic, maternal and environmental variance components for distal spine and body length from statistical models of among clonal line, among subline, and within subline variation, respectively (; ; ). We fitted linear mixed effects models (LME) for each trait separately (i.e. distal spine or body length) using the nlme package () in r version 2.12.0 (). In each case, the trait measured in F2 offspring was modelled by random effects for clonal line and subline nested within clonal line. We assessed the significance of the random effects in two ways: (i) by obtaining 95% confidence intervals around the random effects through bootstrapping () and (ii) through model comparisons using likelihood ratio tests (LRT). 95% confidence intervals were obtained by randomly resampling our data set of distal spine and body lengths (1000 iterations, accounting for clonal and subline structuring), which created a distribution of variances around our random effects. For the LRT, we fitted two additional models for each trait, with each model containing successively fewer random effects. The first additional model for each trait contained the random effect for clonal line but not the random effect for subline. The second additional model for each trait was a linear model without either random effect. As Bythotrephes body size increases during instar development, we included instar as a fixed effect (categorical for instars 1, 2 and 3) in models of body length. The only fixed effect in models of distal spine length was the intercept.We also tested for temporal differences in genetic and maternal variance components by assessing differences by month within 2008 (July, September, and November) and assessing differences across years. We compared a model containing time period (either month or year) as a fixed effect and all random effects described earlier to a model containing the same fixed and random effects, but which allowed clonal and subline variation to differ by time period using the varIdent function in the nlme package (). We used a LRT to assess whether separate estimates of clonal and subline variation for each time period significantly improved the fit of the model. It is noted that this approach differs from simply assessing the significance of time period as a fixed effect, which would assess whether mean phenotypes (i.e. distal spine or body length) differ among time periods. In our analysis, we were testing whether time period affected the clonal line and subline random effects in the model.Descendants of clonal organisms are effectively linkage groups for their entire genotype and, therefore, broad-sense heritability (H2) is the appropriate measure of inheritance (; ). We estimated genetic (Vg), maternal (Vm) and environmental (Ve) variation from the variance among clonal lines, variance among clonal sublines and variance within clonal sublines, respectively, from our mixed-effect model analysis of our clonal breeding design. Based on , we estimated the coefficient of genetic variation (CVg) as: accounting for clonal and subline structure in the trait mean. We used the variance components to calculate H2 as: Similar to the calculation of H2, we calculated maternal effects (m2) as the ratio of maternal variance (Vm) to total phenotypic variance (Vg + Vm + Ve). […]

Pipeline specifications

Software tools lme4, nlme
Application Mathematical modeling