Influence of P. flexilis ancestry on expression patterns and survival
Survival for seedlings was recorded in August 2019 as a binary trait. To obtain maternal tree and population-level estimates of survival, we fitted a similar linear mixed effect model as used for the expression data, but with a binomial error distribution (Swenson 2021), as implemented with the glmmTMB v. 1.1.2.3 package (Brooks et al ., 2017) in R.
To assess the impact of genomic ancestry on adaptive trait differentiation (H3) at the per-transcript level, we correlated mean population-level estimates of P. flexilis ancestry obtained from NGSAdmix (Fig. S2b) with population-level expression values for eachQ ST category. For Cold-condA and Warm-condA, we used expression values estimated at cold and warm gardens, respectively. For Ad-Pl we used the absolute difference in expression values across gardens. For each Q ST category, the observed distribution of Pearson’s correlation coefficients (r ) was compared against an empirically determined background distribution based on 10,000 permutations that were matched to the expression levels of transcripts in the category of interest. Significant deviation in the observed dataset was evaluated using the Kolmogorov-Smirnov test implemented in R (p < 0.05). We report results only from a simple permutation test that does not involve expression level binning since our results were not sensitive to binning and expression levels did not vary much by category (results not shown). Using a similar approach, we evaluated whether the threeQ ST categories deviated significantly from the background set of transcripts in their association with the estimated population-level survival measured at the respective gardens and the mean across gardens for the Ad-Pl category. Similarly, we tested the third hypothesis at the module level by evaluating whether the eigengene expression for a module was associated with both P. flexilisancestry and survival using a false discovery rate-based multiple testing correction as implemented in the stats package in R (q< 0.05).