A Bayesian model for the analysis of transgenerational epigenetic variation.
Bottom Line: The new procedure was used with two simulated data sets and with a beef cattle database.In the simulated populations, the results of the analysis provided marginal posterior distributions that included the population parameters in the regions of highest posterior density.In the case of the beef cattle dataset, the posterior estimate of transgenerational epigenetic variability was very low and a model comparison test indicated that a model that did not included it was the most plausible.
Affiliation: Unidad de Genética Cuantitativa y Mejora Animal, Universidad de Zaragoza, 50013, Zaragoza, Spain Instituto de Biocomputación y Física de los Sistemas Complejos (BIFI), Universidad de Zaragoza, 50018, Zaragoza, Spain email@example.com.Show MeSH
Mentions: The proposed procedure was first checked with simulated data. The summary of the marginal posterior distributions for the variance components, additive genetic (h2), and epigenetic () heritability, reset coefficient (v), and epigenetic transmission coefficient (1 − v) for both cases of simulation are presented in Table 1. The greatest posterior density at 95% for all parameters in the model included the simulated values. However, some details of the results should be highlighted. In the second case of simulation, with lower and λ, the posterior standard deviation (and the HPD95) for the and were remarkable wider (66.19 and 67.42, respectively). The cause of this phenomenon is that there is a statistical confounding between the epigenetic and residual variance components when the reset coefficient (v) is very high, because the T and I matrices becomes very similar. To illustrate this fact, in Figure 2 and Figure 3, we present the joint posterior densities of and v for both cases of simulation. The results of the first case of simulation showed posterior independence between both parameters, whereas in the second, the marginal posterior density presented a half-moon shape. This indicates that for large values of v, (and ) may take any value on its support with a fairly equal probability. In other words, in the first simulated case the model showed very good ability to discriminate between both parameters, whereas in the second it did not. In addition, mixing of the MCMC procedure in the second case of simulation was clearly worst, and adaptive MCMC algorithms, such as the proposed by Mathew et al. (2012) may represent an interesting alternative for its implementation in large data sets.
Affiliation: Unidad de Genética Cuantitativa y Mejora Animal, Universidad de Zaragoza, 50013, Zaragoza, Spain Instituto de Biocomputación y Física de los Sistemas Complejos (BIFI), Universidad de Zaragoza, 50018, Zaragoza, Spain firstname.lastname@example.org.