In stefpeschel/NetCoMi: Network Construction and Comparison for Microbiome Data

knitr::opts_chunk$set(fig.path="figures/tutorial_adapt/")

Tutorial: How to adapt taxa names in the network plot

Use genus names instead of ASV names: Method 1

In this method, genus names are passed as label vector to the plot function, so that genus names are shown in the plot, but not in the summary outputs.

library(NetCoMi)
library(phyloseq)

data("amgut2.filt.phy")

colnames(amgut2.filt.phy@tax_table@.Data) <- c("Kingdom","Phylum","Class",
                                               "Order", "Family", "Genus", 
                                               "Species")

# Agglomerate to genus level
amgut_genus <- tax_glom(amgut2.filt.phy,"Genus")

# Split data set into seasonal allergies = yes/no
levels(phyloseq::get_variable(amgut_genus, "SEASONAL_ALLERGIES"))
amgut_g_yes = subset_samples(amgut_genus, SEASONAL_ALLERGIES == "yes")
amgut_g_no = subset_samples(amgut_genus, SEASONAL_ALLERGIES == "no")


# Network construction
net_season <- netConstruct(data = amgut_g_yes, 
                           data2 = amgut_g_no,  
                           measure = "pearson",
                           normMethod = "clr", 
                           zeroMethod = "none",
                           sparsMethod = "threshold",
                           thresh = 0.4,
                           verbose = 1,
                           seed = 123456)

# Network analysis
props_season <- netAnalyze(net_season, 
                           clustMethod = "cluster_fast_greedy")

summary(props_season, showCentr = c("degree", "eigenvector"))

# Create label vector
labels <- as.vector(tax_table(amgut_genus)[, "Genus"])
names(labels) <- rownames(tax_table(amgut_genus))

# Optional: Shorten labels to a desired length
#labels <- substr(labels, 1, 9)

plot(props_season, 
     repulsion = 0.9,
     labels = labels,
     sameLayout = TRUE, 
     layoutGroup = "union",
     rmSingles = "inboth", 
     shortenLabels = "intelligent",
     labelPattern = c(5,"'",3),
     labelLength = 10,
     nodeSize = "mclr", 
     labelScale = FALSE,
     cexNodes = 1.5, 
     cexLabels = 0.8,
     cexHubLabels = 0.8,
     cexTitle = 1.2,
     groupNames = c("Wet", "Dry"),
     hubBorderCol  = "gray40",
     mar = c(2, 6, 4, 6))

# With shortened node labels:
labels <- substr(labels, 1, 9)

plot(props_season, 
     repulsion = 0.9,
     labels = labels,
     sameLayout = TRUE, 
     layoutGroup = "union",
     rmSingles = "inboth", 
     shortenLabels = "intelligent",
     labelPattern = c(5,"'",3),
     labelLength = 10,
     nodeSize = "mclr", 
     labelScale = FALSE,
     cexNodes = 1.5, 
     cexLabels = 0.8,
     cexHubLabels = 0.8,
     cexTitle = 1.2,
     groupNames = c("Wet", "Dry"),
     hubBorderCol  = "gray40",
     mar = c(2, 6, 4, 6))

Use genus names instead of ASV names: Method 2

In this method, row names of the count table (stored as otu_table in the phyloseq object) are set to genus names.

Advantages:

Genus names are shown in the summaries
The functionality for adapting labels provided by NetCoMi's plot function can be applied

data("amgut2.filt.phy")

colnames(amgut2.filt.phy@tax_table@.Data) <- c("Kingdom","Phylum","Class",
                                               "Order", "Family", "Genus", 
                                               "Species")

# Agglomerate to genus level
amgut_genus <- tax_glom(amgut2.filt.phy,"Genus")

# Split data set into seasonal allergies = yes/no
levels(phyloseq::get_variable(amgut_genus, "SEASONAL_ALLERGIES"))
amgut_g_yes = subset_samples(amgut_genus, SEASONAL_ALLERGIES == "yes")
amgut_g_no = subset_samples(amgut_genus, SEASONAL_ALLERGIES == "no")


# Rename taxa
# (Since NetCoMi uses only the otu table, we just rename it's rownames)
g_names <- as.vector(tax_table(amgut_genus)[, "Genus"])

taxtab <- amgut_genus@tax_table@.Data

# Make labels unique
duplis <- g_names[duplicated(g_names) | duplicated(g_names, fromLast=TRUE)]

while(length(duplis) > 0){
    duplis.sel <- duplis[duplis == duplis[1]]

    g_names[g_names == duplis.sel[1]] <- paste0(duplis.sel[1], 1:length(duplis.sel))

      duplis <- g_names[duplicated(g_names) | duplicated(g_names, fromLast=TRUE)]
}

rownames(amgut_g_yes@otu_table@.Data) <- g_names
rownames(amgut_g_no@otu_table@.Data) <- g_names


# Network construction
net_season <- netConstruct(data = amgut_g_yes, 
                           data2 = amgut_g_no,  
                           measure = "pearson",
                           normMethod = "clr", 
                           zeroMethod = "none",
                           sparsMethod = "threshold",
                           thresh = 0.4,
                           verbose = 1,
                           seed = 123456)

# Network analysis
props_season <- netAnalyze(net_season, 
                           clustMethod = "cluster_fast_greedy")

# Network analysis summary (which now shows genus names)
summary(props_season, showCentr = c("degree", "eigenvector"))


plot(props_season, 
     repulsion = 0.9,
     sameLayout = TRUE, 
     layoutGroup = "union",
     rmSingles = "inboth", 
     shortenLabels = "simple",
     labelLength = 9,
     nodeSize = "mclr", 
     labelScale = FALSE,
     cexNodes = 1.5, 
     cexLabels = 0.8,
     cexHubLabels = 0.8,
     cexTitle = 1.2,
     groupNames = c("Wet", "Dry"),
     hubBorderCol  = "gray40",
     mar = c(2, 6, 4, 6))

Only hub nodes are labeled

If cexLabels is set to zero, only the hub node's labels are plotted.

plot(props_season, 
     repulsion = 0.9,
     sameLayout = TRUE, 
     layoutGroup = "union",
     rmSingles = "inboth", 
     shortenLabels = "simple",
     labelLength = 9,
     nodeSize = "mclr", 
     labelScale = FALSE,
     cexNodes = 1.5, 
     cexLabels = 0,
     cexHubLabels = 1,
     cexTitle = 1.2,
     groupNames = c("Wet", "Dry"),
     hubBorderCol  = "gray40",
     mar = c(2, 6, 4, 6))