Overview

Taxonomic diversity can be measured using a variety of statistical models. Alpha diversity represents the species diversity (richness, evenness, compositional complexity) within experimental samples and treatment a group (e.g. an experiment group) and beta diversity (next section) examines diversity between groups.

'Diversity' is based on the amplicon sequence variants (ASV) detected by DADA2 per sample.

Alpha diversity was estimated In this section we can estimate alpha diversity using Shannon’s index, Simpson's index, Chao1 richness and observed ASVs. Statistical differences between treatment groups for each of the diversity indices was calculated using pairwise Kruskal-Wallis. Species richness is also plotted as rarefaction curves.Multiple diversity indices were used because each Different diversity indices have different strengths and weaknesses and are better contextually - i.e. depending on the structure and nature of your dataset. For example, Shannon’s index is more sensitive to sample size and influenced by rare ASVs. Simpson's index is less sensitive to sample size and or rare ASVs, but is influenced by dominance/abundance of ASVs. The multiple methods are included in this report so the researcher can decide which is most appropriate for their experimental design.

A good overview of most of these methods can be seen in this paper:

...

5. Combined Shannon and Observed ASV. An additional plot combing both Shannon’s index and Observed ASV indices has been included, to compare similarities and differences between these results. As each index uses different units, results for both have been normalised between 0 and 1.

6. Kruskal-Wallis rank sum test is a rank-based nonparametric test that can be used to determine if there are statistically significant differences between two or more groups. This statistical analysis is provided for each plot, to estimate if there is a significant difference (q.value < 0.05) between all groups.

7. Pairwise Wilcoxon rank sum test (AKA: Mann-Whitney test is the same as the Kruskal-Wallis test, but applied pairwise to each group (technically, The Kruskal-Wallis test is the generalization of the Wilcoxon rank sum test).

...

4a. Preparing your data

...

ampvis2 requires 3 components:

1. A table of read counts per sample per ASV (abundance table)

2. A table that matches ASVs to taxonomy information

3. A metadata table containing your sample IDs and any variables (e.g. experimental groups) associated with each sample.

...

nfcore/ampliseq generates the abundance and taxonomy tables, and we created the metadata table in the previous section, so we can import all 3 into R.

Import the metadata table

When you ran your sequences through the ampliseq pipeline, you submitted the samples with a metadata file. This file contains information on your samples and variables. We need to import this metadata file to run our analysis on selected variables.

samples_table <- read.table("../illumina/data/metadata.tsv", sep = "\t", header = T)

Have a look at your samples table and variables (metadata):

samples_table

Import the ASV abundance table

First import the unfiltered ASV table (ampvis2 does internal filtering).

asvtable <- read.table("../illumina/results/dada2/abundance_table/unfiltered/feature-ASV_table.tsv", check.names = FALSE, sep = "\t", skip = 1, stringsAsFactors = FALSE, comment.char = "", header = T)
colnames(asvtable)[1] <- "ASV"

Have a look at the top few rows of the ASV table, to see if it looks right. The sample IDs should be the column names and the ASV IDs in the first column. All the other columns should contain numbers (i.e. the count of the number of times each ASV was found in each sample).

head(asvtable)

Important: always run this next code cell, even if you haven't subset your data If you're working with a subset of your data (see section 3) you'll need to choose a name for your subgroup (e.g. subgroup <- "high_fat") or if you haven't subset your data (i.e. you're working with all your samples) subgroup <- "all" . This is for naming output plots and files by subset group, so you're not overwriting them each time you run a subset of your data.

subgroup <- "all"

Convert the imported data to ampvis2 format

The following code cell manipulates the data in a variety of ways (see the in-code #comments for explanations) to prepare the data for conversion to an ampliseq2 database.

Import the taxonomy table

Read in the taxonomy data - i.e. the taxonomic assignments for each ASV

mytax <- read.table("../illumina/results/taxonomy/taxonomydada2/ASV_tax.silva_138.tsv", sep = "\t", header = T, quote = "")

Each taxonomic level needs to be in a separate column, and in taxonomy.tsv they are all combined and separated by ';' so we need to split by that delimiter. Use tidyr::separate function for this.

...

Have a look again at the top 6 rows

head(mytax)

Combine the taxonomy and ASV tables

asv_table <- cbind(asvtable, mytaxsplitmytax)
# Also remove the Feature.ASV_ID, Confidence and Confidencesequence columns, as they are not needed
asv_table <- subset(asv_table, select=-c(Feature.ASV_ID, Confidence))

...

confidence, sequence))

asv_table$Kingdom <- gsub("D_0", "k", asv_table$Kingdom)
asv_table$Phylum <- gsub("D_1", "p", asv_table$Phylum)
asv_table$Class <- gsub("D_2", "c", asv_table$Class)
asv_table$Order <- gsub("D_3", "o", asv_table$Order)
asv_table$Family <- gsub("D_4", "f", asv_table$Family)
asv_table$Genus <- gsub("D_5", "g", asv_table$Genus)
asv_table$Species <- gsub("D_6", "s", asv_table$Species)head(asv_table)

Convert the imported data to ampvis2 format

Now combine the samples data with the ASV table using amp_load(). This creates an ampvis2 database that can be used by ampvis2

ampvisdata <- amp_load(otutable = asv_table,
              metadata = samples_table)

...

You can see information about the ampvis2 object you just created by typing its name

ampvisdata

4b. Choosing a

...

variable to analyse

In your metadata you'll usually have multiple variables. These need to be analysed individually, by selecting the variable in this section, then running the remaining analysis sections on this chosen variable. You can then re-run the analysis on another variable by returning to this section, changing the variable name, then running again the remaining analysis sections.NOTE This section is for choosing categorical variables only. See section 8 onward for analysis of continous (i.e. numeric) variables.

You can view your variables as column names in your samples_table:

...

Enter the column name of the variable you want to analyse (i.e. change group <- "Myvariable" in the below cell to your chosen variable's column name). This has to be exactly the same as the column name, including capitalisation, characters such as underscores, etc.

group <- "TimeNose_size"

Ordering your variable

The plotting done in ampvis2 is done by the ggplot2 package. ggplot factorises variables and automatically orders them on the plot by alphabetical order.

...

Choose how you want to order your groups here:

lev <- c("Small", "Medium", "Big")

To order your variable you need to put all the variable levels into the lev = c(..). Make sure each level is in double quotes and separated by a comma.

Then run the following to apply the levels to your data:

ampvisdata$metadata[[group]] <- factor(ampvisdata$metadata[[group]], levels = lev)

4c. Rarefaction curve

This section is for plotting rarefaction curves for your samples, coloured by your chosen variable (if you want to change variables, go back and re-run section 4, choosing a different variable).

Use the ampvis2 function amp_rarecurve() to generate the rarefaction curve plot.

p <- amp_rarecurve(ampvisdata, color_by = group, stepsize = 200)

You can print this out as-is simply by:

p

Modifying plot attributes

You can make additional modifications to the plot colours, axis labels, font size, theme, etc:

p <- p + scale_color_manual(values=c("0h"red", "green", "blue", "orange", "black", "forestgreen", "firebrick", "goldenrod", "darkviolet", "2holivedrab", "4hpurple", "6htomato", "8hyellow", "16h", "24h", "32h", "36h", "48h", "60h", "72h", "96h", "120h")

To order your variable you need to put all the variable levels into the lev = c(..). Make sure each level is in double quotes and separated by a comma.

Then run the following to apply the levels to your data:

ampvisdata$metadata[[group]] <- factor(ampvisdata$metadata[[group]], levels = lev)

...

"cornflowerblue")) +
theme_bw() +
ylab("Number of observed ASVs") +
theme(text = element_text(size = 14))
p

You can edit the above modifications and then re-run the code cell to generate a plot of your liking. The modifications in the above code are:

Colours - scale_color_manual(). See the available R colours here: http://www.stat.columbia.edu/~tzheng/files/Rcolor.pdf. If you set colours manually, you need to enter as many colours (or more, never fewer) as there are categories in your variable.

Theme - theme_bw(). There are many themes available, seen here: https://ggplot2.tidyverse.org/reference/ggtheme.html . You can change theme_bw() to theme_minimal() or theme_light() for example, to see how it looks.

Y axis label - ylab(). You can change this to what you like. You can also add an ylab() to change the X axis label.

Font size - theme(text = element_text(size = 14)). There are many things you can change in the theme, such as text position, font type, plot margins, etc. See here: https://ggplot2.tidyverse.org/reference/element.html

There are a multitude of other modifications you can make to a ggplot, too many to describe in this workshop. But there are plenty of online guides on how to modify ggplot plots. Here's an example: http://www.sthda.com/english/wiki/be-awesome-in-ggplot2-a-practical-guide-to-be-highly-effective-r-software-and-data-visualization

The above changes and more can be applied to any ampvis2 or ggplot plot in this workshop.

Once you have your rarefecation plot looking how you like, you can export it as a 300dpi (i.e. publication quality) tiff or pdf file:

Exporting your plot as a file

You can save your plot as a 300dpi (i.e. publication quality) tiff or pdf file. These files can be found in your working directory.

Tip: you can adjust the width and height of the saved images by changing width = and height = in the code below (and every time ggsave appears in this workflow).

Export as a 300dpi TIFF

tiff_exp <- paste0("rarefaction_", group, "_samples_.tiff")
ggsave(file = tiff_exp, dpi = 300, compression = "lzw", device = "tiff", plot = p, width = 20, height = 20, units = "cm")

Export as a pdf

pdf_exp <- paste0("rarefaction_", group, "_samples.pdf")
ggsave(file = pdf_exp, device = "pdf", plot = p, width = 20, height = 20, units = "cm")

You can now find these files in your working directory (which you originally defined in the 'Setting up your analysis environment' section).

4d. Diversity index plots and statistics - single categorical variable

The overview section outlined (with links and references) the alpha diversity indices that can be examined.

Briefly, these are: Shannon’s index, Simpson's index, Chao1 richness and Observed ASVs. Each of these has strengths and weaknesses. It's up to you, the researcher, to explore the literature and decide which is the best index to use for your data.

This section will generate the diversity index scores and then plot them.

This is based on the variable you chose in the previous sections. If you’ve forgotten what variable group you’ve selected, type:

group

Calculate alpha diversity

First you need to calculate the alpha diversity index scores using ampvis2 function amp_alphadiv(). This will calculate all 4 indices.

div_indices <- amp_alphadiv(ampvisdata, richness = T)

The raw diversity index scores are now attached to your metadata table. You can view the raw scores by (scroll to last 5 columns for scores):

div_indices

Export this table to your working directory as a csv file:

write.csv(div_indices, "diversity_indices_raw_scores.csv")

Choose the index you want to plot

Choose the diversity index scores you want to plot.

indicname <- "Shannon"

Shannon's index is used by default ("Shannon") Change this to "Simpson" to plot Simpson's index scores, "Chao1" for Chao1 richness or "ObservedOTUs" for Observed ASVs.

Box and whisker plot

Create a box and whisker plot showing diversity per group. You can view the basic plot like so:

p <- ggplot(div_indices, aes_string(x=group, y=indicname)) + 
  geom_boxplot(fill = "grey", outlier.size = 2.5, lwd=0.7)
p

And you can use the above plot as-is or modify it. For example:

p <- p + theme_bw() + 
ylab(indicname) + 
xlab("") + 
coord_flip() + 
theme(text = element_text(size = 17)) + 
theme(legend.position = "none") + 
geom_smooth(se = F, aes(group=1, color = "red"))
p

Change the above code to modify the plot as you like. See the rarefaction curve section for more details on modifying a plot.

Briefly, the above code changes the theme (theme_bw()), the Y axis label (ylab(indicname) - note: the current y axis name is inside a variable called indicname, so you need to replace it with text in quotes, e.g. "Chao richness"), removes the X axis label (xlab("")), flips the X and Y axis (coord_flip()), changes the font size to 17 and removes the legend (both in theme()). Finally, a trend line is added (geom_smooth()). Any of these components can be removed or modified as you like.

Save your plot as a tiff and pdf file.

tiff_exp <- paste0("alpha_div_box_plot_", group, "_", indicname, "_samples.tiff")
ggsave(file = tiff_exp, dpi = 300, compression = "lzw", device = "tiff", plot = p, width = 20, height = 20, units = "cm")
pdf_exp <- paste0("alpha_div_box_plot_", group, "_", indicname,"_samples.pdf")
ggsave(file = pdf_exp, device = "pdf", plot = p, width = 20, height = 20, units = "cm")

Statistical analysis

To compare the differences between groups within your variable, a Kruskal-Wallis test (one-way analysis of variance) can be performed to test for overall differences between all groups, and a Wilcoxon rank sum to test pairwise differences between each group.

Kruskal-Wallis:

kr_all <- kruskal.test(get(indicname) ~ get(group), data = div_indices)

This runs the test on the diversity index you chose: indicname, the variable: group and uses the diversity indices table you generated: div_indices

To remind you what diversity index and variable you're examining:

paste0("Diverity index: ", indicname)
paste0("Variable: ", group)

To see the results:

kr_all

Pairwise Wilcoxon rank sum test:

wt_pair <- pairwise.wilcox.test(div_indices[[indicname]], div_indices[[group]], p.adjust.method = "BH")
wt_pair <- data.frame(wt_pair$p.value)
colnames(wt_pair) <- gsub("X", "", colnames(wt_pair))

To see the pairwise results (p values).

wt_pair

Go to the next section: Beta Diversity