Computational protocol: Phylogenetic congruence of lichenised fungi and algae is affected by spatial scale and taxonomic diversity

Similar protocols

Protocol publication

[…] For each of the six datasets, algal and fungal DNA sequences were aligned separately using ‘Prankster’ () with the default parameters. These alignments were used to calculate genetic distance matrices (see ) from the raw distances using uncorrected p-distances () implemented by the ‘dist.dna’ function in the ‘ape’ package in R (). We repeated the analyses using a genetic distance matrix calculated using the TN93 substitution model (). We also repeated these analyses using patristic distances derived from a Bayesian phylogenetic analysis to enable us to compare this ‘tree-free’ method to one based on a full phylogenetic analysis. Bayesian trees were calculated using a lognormal molecular clock in BEAST v1.8. We used these results to calculate patristic distances (sum of branch lengths) from the maximum likelihood tree (calculated in MEGA v.6.0) using the R package ‘adephylo’. The results from both alternative analyses (not shown) were congruent with those obtained using p-distances, so we present the p-distance results only.To relate fungal and algal genetic distances to each other and to describe their variation in space, distance matrices were used in a combined analysis using Moran’s eigenvector mapping and variance partitioning (, pp. 258). This analysis was used to describe and partition the variation in algal genetic distances between (a) the fungal genetic distance matrix and (b) a matrix of spatial variables. The spatial variables were derived using the Moran’s eigenvector maps (MEMs) procedure () implemented using the ‘pcnm’ function in the R package ‘vegan’, which uses a principal coordinates analysis to represent different scales of spatial variation for the given set of sample locations (). MEM analysis produces one fewer spatial variables than there are sample points, describing all possible spatial variation in the data from broad scale variation to very fine scale variation. Only the most important subset of these spatial variables (MEMs) was included in a distance-based redundancy analysis (db-RDA) analysis, which relates multivariate data (algal genetic distance matrix) to explanatory matrices (fungal genetic distance matrix and spatial variables). The selected MEMs were those that were significantly related to the algal distance matrix in a distance-based RDA and forward selection procedure using the ‘capscale’ and ‘ordistep’ functions in the ‘vegan’ package in R (). Variance partitioning calculations were conducted following procedures outlined in . The ‘capscale’ function performs a redundancy analysis that seeks the series of linear combinations of the explanatory factors that best describe variation in the response matrix, constrained by the two explanatory matrices (). The variance partitioning procedure computes R2 canonical values analogous to the adjusted R2 values produced in multiple regression (). The analysis indicates how much total variation in the response matrix (e.g., algal genetic distance) is explained by each of the explanatory matrices alone, as well as the component of shared variation, e.g., spatially structured variation in fungal genetic distances. This analysis was also performed using the fungal genetic distance matrix as the dependent matrix and the algal genetic distance matrix as the explanatory matrix to allow comparison of the degree of spatial correlation in each of the matrices.To test the significance of the phylogenetic congruence between fungi and algae for each of the six datasets, we used the Procrustes approach to co-phylogeny (PACo, ). This procedure performs a principal coordinates analysis on the algal genetic distance matrix followed by a Procrustes rotation of the fungal genetic distance matrix, while retaining the information that algae and fungi are paired in particular lichen specimens (). A sum of squares is calculated from the individual residuals for each specimen that represents the lack of fit of the fungal genetic distance matrix onto the principal coordinate analysis result for the algal genetic distance matrix (). The algal–fungal pairing matrix, i.e., which alga is paired with which fungus, is then randomised 10,000 times and the sums of square values recalculated. The observed sum of squares value is then compared to the distribution of values from the randomisations to determine the probability of obtaining the observed result under random expectation (). The magnitude of the residual for each lichen specimen shows its relative lack of fit to a co-diversification pattern. Therefore, for three datasets for which we had additional information on specimen traits, we compared individual residuals among specimens to determine which ones contributed most to the observed association pattern. These three datasets (and trait information) were the Flock Hill community (growth form), Flock Hill Usnea and New Zealand Usnea datasets (apothecia present or absent). Raw genetic and geographic distance matrices are provided in and . […]

Pipeline specifications

Software tools PATRISTIC, MEGA-V, adephylo
Applications Phylogenetics, GWAS
Diseases Genetic Diseases, Inborn