The expected response to selection from the radial four-locus network is, however, slightly sigmoid.
This decision was based on the results from the S TRUCTURE analyses, multi-locus network and species tree analyses.
We used P OFAD V1.03 [ 64] and S PLITST REE V4.0 [ 65] to build a multi-locus network based on the mitochondrial and nuclear data sets.
Moreover, because random fluctuations, most likely a result of genetic drift (given the relatively small population sizes generally implied in artificial selection procedures), observed responses from the four-locus network do not systematically exhibit the slightly sigmoid pattern.
Consequently, although the four-locus network detected in the Virginia chickens explains nearly half of the original difference between lines, other factors contributed about 75% of the phenotypic variability among the informative individuals in the F2 population.
To provide a comprehensive base for the analysis of potentially multi-functional loci and variants, respectively, we compiled two network representations of the information made available by GWA studies: the locus-based "shared locus network" (SLN, Figure 1B) and the variant-based "shared variant network" (SVN, Figure 1C).
The multi-locus network derived from the combined analyses of the mitochondrial and nuclear loci recovered: 1) a clear separation of the montane (albigula) and lowland (debilis and rabai) clades, and 2) a greater extent of genetic diversity within the lowland clade relative to the montane clade.
To visualize the effect of individual QTL pairs and also to estimate the total contribution of the epistatic network, we produced a dataset including only F2 individuals where the genotype for all QTL in the main four locus network affecting body-weight at 56 days of age could be determined with high confidence (n = 727).
This is a clearly realistic scenario as the four-locus networks contributes to half the selection response in the Virginia lines, and thus other genetic factors need to contribute as well.
Furthermore, FastEpistasis could detect, with higher than 50% accuracy, in fewer than 50% of the 4-loci network models and only Ep1, Ep3 and Ep5 of the 5-loci network models.
We present a software package for expression quantitative trait loci (eQTL) network analysis and demonstrate how network analysis of whole tissues provides insights into interactions between cell compartments and signaling molecules.
