upset_vp {UpSetVP}

Visualization of VPA and HP Using UpSetVP Diagram

Description

Visualization of variance partitioning analysis (VPA) and hierarchical partitioning (HP) with unlimited number of predictor variables (or matrices of predictors) using UpSet matrix layout.

Usage

upset_vp(
  x,
  plot.hp = TRUE,
  order.part = "effect",
  decreasing.part = TRUE,
  order.var = TRUE,
  decreasing.var = TRUE,
  cutoff = -1,
  nVar = 30,
  col.width = 0.6,
  pch.size = 3,
  line.lwd = 0.5,
  show.effect = TRUE,
  effect.cex = 2.7,
  title.cex = 10,
  axis.cex = 8,
  height.ratio = c(2, 1),
  width.ratio = c(1, 3)
)

Arguments

x

A rdacca.hp object, which contains the output of VPA and HP from canonical analysis.

plot.hp

The default is TRUE, which plots the individual effect for each predictor on left column diagram. If FALSE, compute and plot the sum of unique effect and common effect for each predictor.

order.part

How the VPA components in matrix layout should be ordered. Options include "effect" (order the intersections by their effects) or "degree" (sort by the number of predictors involved in the intersection), default is "effect".

decreasing.part

How the intersections in order.part should be ordered. Default is TRUE, "effect" is decreasing (from greatest to least) or "degree" is increasing (from least to greatest).

order.var

The predictors in the matrix layout should be ordered by. Default is TRUE, which orders the predictors by their effects. IF FALSE, sort by the order of predictors in input data.

decreasing.var

If order.var=TRUE, how the predictors should be ordered. Default is TRUE, from greatest to least.

cutoff

Effects below cutoff will not be displayed, default is -1. Note: Negative effects due to adjustment of R-squared mean negligible contributions, but they are included in the computation of the total contribution of each predictor category.

nVar

Number of components in VPA to plot, default is 30.

col.width

Width of bars in column diagram, default is 0.6.

pch.size

Size of points in matrix diagram, default is 3.

line.lwd

Width of lines in matrix diagram, default is 0.5.

show.effect

Show the relative importance of predictors (unique, common, or individual effects) above bars, default is TRUE.

effect.cex

Font size of the effects, default is 2.7.

title.cex

Font size of axis titles, default is 10.

axis.cex

Font size of axis labels, default is 8.

height.ratio

Ratio between matrix and top column diagram, default is c(2, 1).

width.ratio

Ratio between matrix and left column diagram, default is c(1, 3).

Details

UpSetVP diagram is an extension of UpSet technique (Conway et al. 2017; Lex et al. 2014) to canonical analysis and is used to visualize the object of rdacca.hp (Lai et al. 2022). The matrix layout enables the effective representation of relative importance of predictors, such as the unique effects and common effects in VPA, as well as additional summary statistics or individual effects in HP. UpSetVP diagram could, in principle, allow visualization of any number of predictor variables or groups of predictor variables. But considering the interpretability of data, we would like to recommend that the number of predictors (or groups of predictors) no more than 7.

Value

Returns a ggplot2.

References

Conway J R, Lex A, Gehlenborg N. UpSetR: an R package for the visualization of intersecting sets and their properties. Bioinformatics, 2017, 33(18): 2938-2940.

Gong S, Feng B, Jian S P, et al. Elevation Matters More than Season in Shaping the Heterogeneity of Soil and Root Associated Ectomycorrhizal Fungal Community. Microbiology spectrum, 2022, 10(1): e01950-21.

Lai J, Zou Y, Zhang J, et al. Generalizing hierarchical and variation partitioning in multiple regression and canonical analyses using the rdacca.hp R package. Methods in Ecology and Evolution, 2022.

Lex A, Gehlenborg N, Strobelt H, et al. UpSet: visualization of intersecting sets. IEEE transactions on visualization and computer graphics, 2014, 20(12): 1983-1992.

Examples

library(rdacca.hp)

## A simple example of partial dbRDA
data(baima.fun)
data(baima.env)

# Bray-Curtis index was used to calculate community composition dissimilarity
baima.fun.bray <- vegdist(baima.fun, method = "bray")

# The relative importance of individual soil properties on EcM fungal community compositionon
soil <- baima.env[c("pH", "TP", "TK", "AN", "AP", "AK")]
baima.soil.vp <- rdacca.hp(baima.fun.bray, soil,
  method = "dbRDA", var.part = TRUE, type = "adjR2")

# Plot unique, common, as well as individual effects
upset_vp(baima.soil.vp, plot.hp = TRUE)

## Example was referenced from Gong et al. (2022)
if(requireNamespace("adespatial", quietly = TRUE)) {

# Distance-based Moran's eigenvector maps (dbMEM) was used to extract spatial relationships
space.dbmem <- adespatial::dbmem(baima.env[c("latitude", "lontitude")])
  
  # The relative importance of groups of environmental factors on EcM fungal community composition
  env.list <- list(
    elevation = baima.env["altitude"],
    season = baima.env["season"],
    space = data.frame(space.dbmem)[1:2],
    host = baima.env[c("em.GR", "em.abun")],
    climate = baima.env["sea.MT"],
    soil = baima.env[c("pH", "TP", "TK", "AN", "AP", "AK")]
  )
  baima.env.vp <- rdacca.hp(baima.fun.bray, env.list,
    method = "dbRDA", var.part = TRUE, type = "adjR2")
  
  # Plot unique, common, as well as individual effects
  upset_vp(baima.env.vp, plot.hp = TRUE, order.part = "degree")
}