Computational protocol: Multivariate ordination identifies vegetation types associated with spider conservation in brassica crops

Similar protocols

Protocol publication

[…] To test the importance of vegetation types on spider assemblages in brassica fields and the influence of those habitats on the spatial distribution of spider species, we applied variance partitioning, hierarchical clustering (for community similarities or dissimilarities) and spatial eigenvector analysis for spider abundance and diversity data. Abundance “n” and Shannon-Wiener index “H” () were calculated using the vegan package (vegan 2.4-0) (), in R statistical software (R version 3.4.0), then the data were Hellinger transformed to obtain normality and adjust variance prior to multivariate analysis. The Hellinger transformation has good statistical properties to test for relationships among explanatory variables and draw biplots in constrained or unconstrained multivariate ordination (e.g., redundancy analysis RDA) without resorting to the Euclidean distances () and is also suited to data sets with multiple zero values. We identified the response of spider abundance and diversity (H) against different vegetation types and weighted principal coordinates of neighbor matrices (PCNM) as explanatory variables using the “varpart” and “pcnm” functions of package “vegan” (version 2.4-1) () in R (version 3.4.0), which allowed variance partitioning to separate the effects of weighted PCNM and vegetation types on spider abundance and diversity (H) (). PCNM, also known as Moran’s Eigenvector Maps (MEM), is a powerful approach able to detect spatial or temporal patterns (henceforth, only spatial patterns will be discussed) of varying scale in response data (spider abundance and diversity) (; ; ). Essentially, spatial variables are used to determine the distance between sites with special focus on neighbouring sites. Additionally, the “rda” function of package “vegan” (version 2.4-1) was used to test the significance of fractions of each spider family’s abundance and diversity (H), and triplots were constructed to visualize the vegetation types associated with different spider families. All analyses was carried out separately for each of the three experimental sites because of differences in adjacent vegetation types to the brassica field.To measure community dissimilarities of spiders in different vegetation types, hierarchical clustering was carried out for the abundance and diversity (H) per sampling points at each experimental site. A quantitative version of the Sørensen index, Bray-Curtis dissimilarity was used to measure the percentage differences and to construct dissimilarity matrices for abundance and diversity (H) of spider families in brassica and adjacent crop and non-crop habitat types using the “vegdist” function with “method = “bray”” (; ) using “vegan” (version 2.4-1) (). We visualized the β-dissimilarity matrix using heatmap for the abundance and diversity (H) of spider families at each of the experimental sites (; ; ) by using the “gplots” (), “Heatplus” (), “RColorBrewer” () and “ComplexHeatmap” () packages in R (version 3.4.0). An assessment of the uncertainty in the cluster delineation was done through multiscale nonparametric bootstrap resampling tests () using “pvclust” () package in R (version 3.4.0). This helps to determine p-values (two types: approximately unbiased (AU) p-value and bootstrap probability (BP) value) of each cluster in the hierarchy ().Spatial eigenvector analysis is particularly well suited to data with low spatial or temporal replication, when compared to classical geostatistical analysis (e.g., semivariograms) (; ), which was the case in our data. We were interested in calculating and mapping the spatial variation in the occurrence of spiders, and analyzing its relationship with the adjacent vegetation of the focal brassica field. Distance-based MEM (dbMEM) (; ) was used to control for spatial autocorrelation in tests of abundance and diversity (H) of spider-vegetation relationships, see using the packages “adespatial” (), “ade4” (), “adegraphics” () in R (version 3.4.0). We identified a total of 11 distance based Moran’s eigenvector maps for Minqing, seven for Nantong 1 and nine for Nantong 2. Significant Moran’s eigenvector maps for each of the experimental sites were identified with forward selection using double stop criterion (), α = 0.05 and R2 values (for abundance; R2 = 0.45 in Minqing, R2 = 0.37 in Nantong 1 and R2 = 0.34 in Nantong 2, and for diversity (H); R2 = 0.46 in Minqing, R2 = 0.34 in Nantong 1 and R2 = 0.23 in Nantong 2). We identified one significant Moran’s eigenvector map for spider abundance out of a total of 11 in Minqing and nine for Nantong 2. Whilst for diversity (H); we identified two significant Moran’s eigenvector maps out of a total of 11 in Minqing and one out of nine for Nantong 2. Further, canonical analysis (rda) was performed to compute the dbMEM spatial models and the “anova” function was used to test the significance of these models. All spatial models were found to be highly significant (p-value < 0.001). R-codes and datasets are attached as –. […]

Pipeline specifications

Software tools gplots, Complex heatmaps, Pvclust
Application Transcriptome data visualization
Organisms Colocasia esculenta