Combining Genotype, Phenotype, and Environment to Infer Potential Candidate Genes

Abstract

Population genomic analysis can be an important tool in understanding local adaptation. Identification of potential adaptive loci in such analyses is usually based on the survey of a large genomic dataset in combination with environmental variables. Phenotypic data are less commonly incorporated into such studies, although combining a genome scan analysis with a phenotypic trait analysis can greatly improve the insights obtained from each analysis individually. Here, we aimed to identify loci potentially involved in adaptation to climate in 283 Loblolly pine (Pinus taeda) samples from throughout the species' range in the southeastern United States. We analyzed associations between phenotypic, molecular, and environmental variables from datasets of 3082 single nucleotide polymorphism (SNP) loci and 3 categories of phenotypic traits (gene expression, metabolites, and whole-plant traits). We found only 6 SNP loci that displayed potential signals of local adaptation. Five of the 6 identified SNPs are linked to gene expression traits for lignin development, and 1 is linked with whole-plant traits. We subsequently compared the 6 candidate genes with environmental variables and found a high correlation in only 3 of them (R 2 > 0.2). Our study highlights the need for a combination of genotypes, phenotypes, and environmental variables, and for an appropriate sampling scheme and study design, to improve confidence in the identification of potential candidate genes.

Partitioning genetic and species diversity refines our understanding of species–genetic diversity relationships

“…For our second approach, we identified SNP loci associated with environmental variables using redundancy analysis (RDA), a genome–environment association (GEA) approach to distinguishing candidate loci under selection based on correlation between genotype and environmental factors expected to impose natural selection. RDA is a canonical ordination technique where, first, response variables (multiple loci) are modeled as a function of linear combinations of the predictors (multiple environmental variables), then a PCA of the fitted values produces the RDA components that best explain, in sequential order, the variation among the fitted genetic values (Forester, Jones, Joost, Landguth, & Lasky, ; Legendre & Legendre, ; Talbot et al., ). To check that the final model did not suffer multi‐collinearity problems, we calculated the variance inflation factors (VIFs) and verified that none of them exceeded 5 for any of the predictors.…”

“…For our second approach, we identified SNP loci associated with environmental variables using redundancy analysis ( produces the RDA components that best explain, in sequential order, the variation among the fitted genetic values (Forester, Jones, Joost, Landguth, & Lasky, 2016;Legendre & Legendre, 2012;Talbot et al, 2017). To check that the final model did not suffer multi-collinearity problems, we calculated the variance inflation factors (VIFs) and verified that none of them exceeded 5 for any of the predictors.…”

Detecting the genetic basis of local adaptation in loblolly pine (Pinus taedaL.) using whole exome‐wide genotyping and an integrative landscape genomics analysis approach

“…Talbot et al (2017) reported that loci with local adaptation signatures in loblolly pine were also linked to gene expression traits for lignin development and wholeplant traits. Talbot et al (2017) reported that loci with local adaptation signatures in loblolly pine were also linked to gene expression traits for lignin development and wholeplant traits.…”

Genomic footprints of adaptation in a cooperatively breeding tropical bird across a vegetation gradient

“…After removing two layers (Bio1, annual mean temperature and Bio7, temperature annual range = Bio5–Bio6) that showed collinearity with other variables, we evaluated which of the remaining layers best predicted phenotypic variation through a series of generalized linear models (GLMs) with stepwise (backward) logistic regressions. These models were used as guidelines to identify the variants that could be influencing phenotypic evolution in the Yucatan jay, following a similar rationale as in previous studies in humans (Hancock et al., ) and pine (Talbot et al., ). Each morphometric trait was independently correlated to the remaining 17 bioclimatic variables, and the model that best explained each trait was selected based on the Akaike information criterion.…”

“…Furthermore, the availability of bird specimens in collections makes the assessment of morphometric variation easier. When correctly integrated into genotype–environment associations, this variation can greatly improve the identification of candidate genes and connect genomics to fitness‐related traits (e.g., Haasl & Payseur, ; Hancock et al., ; Talbot et al., ).…”