Pseudobulk analysis is a powerful technique that bridges the gap between single-cell and bulk RNA-seq data. It involves aggregating gene expression data from groups of cells within the same biological replicate, such as a mouse or patient, typically based on clustering or cell type annotations (Murphy and Skene 2022).

A key advantage of this approach in differential expression (DE) analysis is that it avoids treating individual cells as independent samples, which can underestimate variance and lead to inflated significance or overly optimistic p-values (Squair et al. 2021). This occurs because cells from the same biological replicate are inherently more similar to each other than cells from different samples. By grouping data into pseudobulk samples, the analysis aligns with the experimental design, as in bulk RNA-seq, leading to more reliable and robust statistical results (Murphy and Skene 2022).

Beyond enhancing statistical validity, pseudobulk analysis enables the identification of cell-type-specific gene expression and functional changes across biological conditions. It balances the detailed resolution of single-cell data with the statistical power of bulk RNA-seq, providing insights into the functional transcriptomic landscape relevant to biological questions. Overall, for DE analysis in multi-sample single-cell experiments, pseudobulk approaches demonstrate superior performance compared to single-cell-specific DE methods (Squair et al. 2021).

In this tutorial, we will guide you through a pseudobulk analysis workflow using the Decoupler (Mompel et al. 2022) and edgeR (Liu et al. 2015) tools available in Galaxy. These tools facilitate functional and differential expression analysis, and their output can be integrated with other Galaxy tools to visualize results, such as creating Volcano Plots, which we will also cover in this tutorial.

Agenda

In this tutorial, we will cover:

1. Let’s Get the Data!
   1. Overview of the Data
   2. Data Upload
2. Generation of the Pseudobulk Count Matrix with Decoupler
   1. Sanitation Steps - Part 1
   2. Generating the Contrast File
3. Differential Expression Analysis (DE) with edgeR
   1. Sanitation Steps - Part 2
4. Volcano Plot
   1. What are the results of the Volcano Plot?
5. Subsetting Samples from the Original AnnData Object
   1. Extracting observations of interest as AnnData object
   2. Recommendations

Let’s Get the Data!

Overview of the Data

Our data was extracted from the publication titled “Elevated Calprotectin and Abnormal Myeloid Cell Subsets Discriminate Severe from Mild COVID-19” (Silvin et al. 2020). This dataset was chosen because it was utilized by the developers of the decoupler tool for pseudobulk aggregate analysis (Lab and Contributors).

Pseudobulk analysis is an advanced method in single-cell data analysis. For this tutorial, we assume familiarity with common single-cell data formats, such as AnnData or Seurat objects, and experience analysing single-cell data, including clustering and annotating cell types.

If you’re new to these concepts, we recommend exploring our other tutorials before performing pseudobulk analysis:

• Clustering 3K PBMCs with Scanpy: Learn how to cluster and annotate cells in Galaxy using our single-cell tools.
• Combining single cell datasets after pre-processing: Understand how to combine multiple datasets into one AnnData object and add metadata from single-cell experiments.

The data object, which you will import from Zenodo into Galaxy via the provided link, has been preprocessed, analysed, and annotated. It includes the following key observation metadata:

• cell_type: The type of cell identified.
• disease: Indicates whether the sample is a control or corresponds to a COVID-19 condition.
• individual: The individual donor or sample identifier.
• sex: The sex of the individual.

Data Upload

Hands On: Upload your data

1. Create a new history for this tutorial and name it “Pseudobulk DE Analysis with edgeR”

2. Import the AnnData file from Zenodo:

   https://zenodo.org/records/13929549/files/Source_AnnData_file.h5ad
   

   Tip: Importing via links

   • Copy the link location

   • Click galaxy-upload Upload Data at the top of the tool panel

   • Select galaxy-wf-edit Paste/Fetch Data

   • Paste the link(s) into the text field

   • Press Start

   • Close the window

   Tip: Importing data from a data library

   As an alternative to uploading the data from a URL or your computer, the files may also have been made available from a shared data library:

   1. Go into Libraries (left panel)
   2. Navigate to the correct folder as indicated by your instructor.
      • On most Galaxies tutorial data will be provided in a folder named GTN - Material –> Topic Name -> Tutorial Name.
   3. Select the desired files
   4. Click on Add to History galaxy-dropdown near the top and select as Datasets from the dropdown menu

   5. In the pop-up window, choose

      • “Select history”: the history you want to import the data to (or create a new one)
   6. Click on Import

3. Rename the dataset: “AnnData for Pseudobulk.”

4. Ensure that the datatype is correct. It should be an AnnData object (h5ad).

   Tip: Changing the datatype

   • Click on the galaxy-pencil pencil icon for the dataset to edit its attributes
   • In the central panel, click galaxy-chart-select-data Datatypes tab on the top
   • In the galaxy-chart-select-data Assign Datatype, select datatypes from “New type” dropdown
     • Tip: you can start typing the datatype into the field to filter the dropdown menu
   • Click the Save button

Generation of the Pseudobulk Count Matrix with Decoupler

In this step, our goal is to perform a “bioinformatic cell sorting” based on the annotated clusters of the single-cell data.

To start a pseudobulk analysis, ensure that the AnnData object you will use contains all the necessary metadata for “pseudobulking”. For example, key annotations, such as cell_type, condition, disease, and batch, should be present. Most importantly, the AnnData object has to include a layer with the raw counts of gene expression.

Raw counts are crucial for generating accurate pseudobulk aggregates. Since single-cell data is typically normalized after annotation, it’s important to preserve the raw counts in the AnnData object before normalization steps if you would like to perform a pseudobulk analysis later on. These raw counts are directly used by tools like Decoupler to generate the pseudobulk count matrix. Note that normalized count matrices should not be used with Decoupler, even if the tool appears to process them successfully.

Tip: Missing Raw Counts?

If your AnnData object lacks raw counts, you can use the AnnData Operations tool on the Single Cell Galaxy instance. This tool allows you to copy the .X matrix into a new layer that you may assign a new label, for e.g., counts or raw_counts, making it available as a parameter for Decoupler.

Importantly, ensure that the copying of this matrix as a raw counts layer is done carefully and correctly. To verify that your AnnData object contains the necessary raw counts layer, you can use the Inspect AnnData Object tool. This tool helps confirm the presence of raw counts and other essential metadata in your AnnData object before proceeding with pseudobulk analysis.

Hands On: Decoupler Pseudobulk

1. Decoupler pseudo-bulk ( Galaxy version 1.4.0+galaxy8) tool with the following parameters:
   • param-file Input AnnData file: AnnData for Pseudobulk (Input dataset obtained > from Zenodo)
   • “Produce a list of genes to filter out per contrast?”: No
   • “Obs Fields to Merge”: (Leave empty if not applicable)
   • “Groupby column”: cell_type (Column containing cell type annotations)
   • “Sample Key column”: individual (Column containing individual sample identifiers)
   • “Layer”: counts (Layer containing raw gene expression counts)
   • “Factor Fields”: disease (Column in adata.obs specifying experimental factors. For edgeR, the first field should be the main contrast field, followed by covariates.)
   • “Use Raw”: No
   • “Produce AnnData with Pseudo-bulk”: (Optional, if yes, it will generate an h5ad output file with pseudobulks)
   • “Minimum Cells:: 10
   • “Produce plots:: Yes
   • “Minimum Counts”: 10
   • “Minimum Total Counts”: 1000
   • “Minimum Counts Per Gene Per Contrast Field”: 20 (Genes with fewer counts in specific contrasts are flagged in a separate file but not excluded from the results.)
   • “Enable Filtering by Expression”: Yes
   • “Plot Samples Figsize”: 13 13
   • “Plot Filtering Figsize”: 13 13

   Comment: Performing DEG within Clusters

   Important! The count matrix retrieved from this tool includes all of our samples individual aggregated by cell_type, which can be identified by the column headers. If you want to perform comparisons of conditions within clusters of each individual cell type, you will need to subset the relevant columns of the matrix and use them as your new count matrix. We will demonstrate this in the last section of this tutorial with detailed hands-on steps.

Question

1. How many outputs does the Decoupler tool generate?
2. What do the outputs represent? What is the interpretation of the plots?
3. Which output(s) will we use in edgeR for differential expression analysis?

Solution

1. The Decoupler tool generates multiple outputs, including:
   • Pseudobulk Count Matrix (tabular file)
   • Samples Metadata (Factor file) (tabular file)
   • Genes Metadata (tabular file)
   • Pseudobulk Plot (PNG format)
   • Filter by Expression Plot (PNG format)
   • Genes to Ignore by Contrast Field (tabular file)
   • Pseudobulk AnnData file (If chosen, h5ad file)
2. The output files contain the following:
   • Pseudobulk Count Matrix: Contains the raw count aggregates for each pseudobulk sample.
   • Samples Metadata (Factor File): Provides metadata annotated for each sample, including factors or annotations added to the AnnData object.
   • Genes Metadata: Includes gene-related information such as gene symbols, Ensembl IDs, dispersion values, etc.
   • Genes to Ignore: Lists of genes that could be excluded or should be carefully considered for specific contrasts. This file contains a contrast field and the corresponding genes written as gene symbols.
   • Pseudobulk Plot: A visual representation of the pseudobulk data.
   • Filter by Expression Plot: Illustrates the expression filtering applied to the data.
   • Pseudobulk AnnData file: An AnnData file that contains the aggregated pseudobulks.
3. The pseudobulk count matrix is the primary input required for analysis using edgeR, a tool designed for differential expression analysis. The Samples Metadata is another file that will serve as an input for the edgeR tool.

Sanitation Steps - Part 1

The next steps will help you refine your data for easier handling. We will use standard galaxy tools, like: Replace Text, Remove columns and Text reformatting.

Hands On: Replace Text and Remove Columns

1. Replace Text ( Galaxy version 9.3+galaxy1) with the following parameters:
   • param-file “File to process”: count_matrix (output of Decoupler pseudo-bulk tool)
   • In “Replacement”:
     • param-repeat “Insert Replacement”
       • “Find pattern”: [ --+*^]+
       • “Replace with:”: _
2. Remove columns ( Galaxy version 1.0) with the following parameters:
   • param-file “Tabular file”: genes_metadata (output of Decoupler pseudo-bulk tool)
   • In “Select Columns”:
     • param-repeat “Insert Select Columns”
       • “Header name”: start
     • param-repeat “Insert Select Columns”
       • “Header name”: end
     • param-repeat “Insert Select Columns”
       • “Header name”: width
3. Replace Text ( Galaxy version 9.3+galaxy1) with the following parameters:
   • param-file “File to process”: samples_metadata (output of Decoupler pseudo-bulk tool)
   • In “Replacement”:
     • param-repeat “Insert Replacement”
       • “Find pattern”: [ --+*^]+
       • “Replace with:”: _
4. Replace Text ( Galaxy version 9.3+galaxy1) with the following parameters:
   • param-file “File to process”: outfile (output of Replace Text: Sample Metadata Step tool)
   • In “Replacement”:
     • param-repeat “Insert Replacement”
       • “in column”: c2
       • “Find pattern”: ^([0-9])(.+)
       • “Replace with”: GG_\\1\\2

In the previous steps, the following modifications were made to the files:

1. Replace Text: Replaces special characters ([ --+*^]+) with underscores (_) in the count_matrix file.
2. Remove Columns: Removes the start, end, and width columns from the genes_metadata file.
3. Replace Text: Replaces special characters ([ --+*^]+) with underscores (_) in the samples_metadata file.

4. Replace Text in Column: Modifies values in column c2 of outfile from step 3 by prefixing GG_ to numbers at the beginning of the string.

   Tip: Regular Expressions 101

   Regular expressions are a standardized way of describing patterns in textual data. They can be extremely useful for tasks such as finding and replacing data. They can be a bit tricky to master, but learning even just a few of the basics can help you get the most out of Galaxy.

   Finding

   Below are just a few examples of basic expressions:

   Regular expression	Matches
   abc	an occurrence of abc within your data
   (abc|def)	abc or def
   [abc]	a single character which is either a, b, or c
   [^abc]	a character that is NOT a, b, nor c
   [a-z]	any lowercase letter
   [a-zA-Z]	any letter (upper or lower case)
   [0-9]	numbers 0-9
   \d	any digit (same as [0-9])
   \D	any non-digit character
   \w	any alphanumeric character
   \W	any non-alphanumeric character
   \s	any whitespace
   \S	any non-whitespace character
   .	any character
   \.
   {x,y}	between x and y repetitions
   ^	the beginning of the line
   $	the end of the line

   Note: you see that characters such as *, ?, ., + etc have a special meaning in a regular expression. If you want to match on those characters, you can escape them with a backslash. So \? matches the question mark character exactly.

   Examples

   Regular expression	matches
   \d{4}	4 digits (e.g. a year)
   chr\d{1,2}	chr followed by 1 or 2 digits
   .*abc$	anything with abc at the end of the line
   ^$	empty line
   ^>.*	Line starting with > (e.g. Fasta header)
   ^[^>].*	Line not starting with > (e.g. Fasta sequence)

   Replacing

   Sometimes you need to capture the exact value you matched on, in order to use it in your replacement, we do this using capture groups (...), which we can refer to using \1, \2 etc for the first and second captured values. If you want to refer to the whole match, use &.

   Regular expression	Input	Captures
   chr(\d{1,2})	chr14	\1 = 14
   (\d{2}) July (\d{4})	24 July 1984	\1 = 24, \2 = 1984

   An expression like s/find/replacement/g indicates a replacement expression, this will search (s) for any occurrence of find, and replace it with replacement. It will do this globally (g) which means it doesn’t stop after the first match.

   Example: s/chr(\d{1,2})/CHR\1/g will replace chr14 with CHR14 etc.

   You can also use replacement modifier such as convert to lower case \L or upper case \U. Example: s/.*/\U&/g will convert the whole text to upper case.

   Note: In Galaxy, you are often asked to provide the find and replacement expressions separately, so you don’t have to use the s/../../g structure.

   There is a lot more you can do with regular expressions, and there are a few different flavours in different tools/programming languages, but these are the most important basics that will already allow you to do many of the tasks you might need in your analysis.

   Tip: RegexOne is a nice interactive tutorial to learn the basics of regular expressions.

   Tip: Regex101.com is a great resource for interactively testing and constructing your regular expressions, it even provides an explanation of a regular expression if you provide one.

   Tip: Cyrilex is a visual regular expression tester.

Generating the Contrast File

This file will be used as the contrast input file in the edgeR tool.

Hands On: Creating a Contrast File for edgeR

1. Text Reformatting ( Galaxy version 9.3+galaxy1) with the following parameters:
   • param-file “File to process”: outfile (output generated from the Replace Text tool step).
   • “AWK Program”:

     BEGIN { print header }
     NR > 1 { if (!seen[$2]++) words[++count]=$2 }
     END { for (i=1; i<=count; i++)
              for (j=i+1; j<=count; j++)
                 print words[i]-words[j] }
     

   Comment: Explanation of the AWK Program

   This AWK script performs the following:

   • Initializes the header variable for the output file.
   • Processes the input file (NR > 1 skips the header line).
   • Tracks unique elements in column 2 (seen[$2]++) to avoid duplicates.
   • Creates pairs of unique elements and calculates their difference, outputting the contrast for downstream edgeR analysis.

Question

1. How does your contrast file look, and what contrast is generated by the tool?

Solution

1. The contrast file is a simple tab-delimited text file. It contains:
   • A header in the first row that labels the column.
   • The contrast, such as normal-COVID_19, written in the second row.

When working with your own data, the contrast file will look different, as it will reflect the specific contrasts in your dataset but the header should be the same

Differential Expression Analysis (DE) with edgeR

The tool edgeR is commonly used for differential expression (DE) analysis of pseudobulk aggregates in single-cell RNA-seq studies (Chen et al. 2024). It calculates gene dispersions based on the total counts for each gene and adjusts these estimates using an approach called empirical Bayes. This method improves reliability by pooling information across all genes in the dataset to stabilize dispersion estimates, particularly for genes with low counts or limited data. By “borrowing strength” from the behavior of other genes, empirical Bayes reduces the risk of overestimating or underestimating variability, ensuring more robust and accurate results (Chen et al. 2024). Differential expression is then evaluated using an exact test for the negative binomial distribution, which is similar to Fisher’s exact test but modified to handle the variability typically seen in RNA-seq data (Chen et al. 2024).

In addition to exact tests, edgeR employs generalized linear models (GLMs), which allow for the analysis of more complex experimental designs. GLMs model gene counts as a function of experimental conditions (e.g., treatment groups or time points) and estimate how these conditions influence gene expression. Likelihood-based methods, such as quasi-likelihood (QL) approaches, are central to this framework. Standard likelihood methods evaluate the fit of the model to the data, while quasi-likelihood methods add an extra layer by explicitly accounting for biological and technical variability, improving the reliability of DE results. These models allow edgeR to identify subtle expression differences while controlling for overdispersion in the data (Chen and Smyth 2016).

Several plots can be generated to assist in understanding the data and the results of the analysis, including Multidimensional scaling (MDS), Biological Coefficient of Variation plot (BCV), Quasi-Likelihood (QL), and Mean-Difference plot o MA (MD) plots. These visualizations provide insights into sample relationships, variability, and differential expression, and will be explained further in the tutorial. With these concepts in mind, let’s now perform our DE analysis using our edgeR tool in Galaxy!

Hands On: Run a DGE Analysis with edgeR

1. edgeR ( Galaxy version 3.36.0+galaxy5) with the following parameters:
   • “Count Files or Matrix?”: Single Count Matrix
     • param-file “Count Matrix”: outfile (output of Replace Text: Count Matrix tool)
     • “Input factor information from file?”: Yes
       • param-file “Factor File”: outfile (output of Replace Text: Creating Factor File tool)
   • “Use Gene Annotations?”: Yes
     • param-file “Gene Annotations”: output_tabular (output of Remove Columns: Gene Metadata tool)
   • “Formula for linear model”: ~ 0 + factor_A or ~ 0 + factor_A + factor_B:factor_C (Customize this formula based on your data’s contrast names and factor file. Ensure the formula matches EdgeR’s syntax and uses only elements from the factor file.)
   • “Input contrasts manually or through a file?”: file
     • param-file “Contrasts File”: outfile (output of Text Reformatting: Creating a Contrast File for edgeR tool)
   • In “Filter Low Counts”:
     • “Filter lowly expressed genes?”: No

   Comment: edgeR Tool Overview

   Use this tool to perform DGE analysis using edgeR. It is highly customizable, allowing you to input your count matrix, factor information, and gene annotations prepared during earlier steps. The contrast file should be generated from the previous step, Text Reformatting: Creating a Contrast File for edgeR. The model formula should be defined based on your dataset and must include factors that were already specified in the contrast file.

Question

1. What are the output(s) of the edgeR tool?
2. How can we interpret our output result file?

Solution

1. Output files:
   • edgeR Tables: This file includes columns with information such as:
   • Gene Symbol: The identifier for each gene.
   • Log Fold Change (logFC): Represents the direction and magnitude of differential expression. Positive values indicate upregulation, and negative values indicate downregulation.
   • Raw p-value (PValue): The unadjusted statistical significance value for each gene, indicating the likelihood of observing the result under the null hypothesis.
   • False Discovery Rate (FDR): The adjusted p-value using the Benjamini-Hochberg correction method to account for multiple testing and control the expected proportion of false positives.
   • (Optional): If additional outputs, such as “normalized counts,” are selected in the edgeR tool, the table will also include these counts.
   • edgeR Report: This is an HTML file containing various visualizations and summaries of the analysis, such as:
   • MDS Plot (Multidimensional Scaling): Visualizes the relationships between samples based on gene expression profiles.
   • BCV Plot (Biological Coefficient of Variation): Displays the estimated dispersion for each gene.
   • QL Plot (Quasi-Likelihood): Shows the quasi-likelihood dispersions for the dataset.
   • MD Plot (Mean-Difference Plot): Compares the mean expression of genes versus their log fold change.
2. Interpreting the Results:
   • The table contains the results of the differential expression analysis.
   • The first column typically lists the gene symbols from the dataset.
   • The logFC indicates the direction and magnitude of differential expression:
   • Upregulated genes: Genes with higher expression in the first group of the contrast (e.g., ‘normal’ group in a “normal-COVID_19” contrast).
   • Downregulated genes: Genes with lower expression in the first group of the contrast (e.g., ‘normal’ group in a “normal-COVID_19” contrast).
   • The raw p-value represents the statistical significance of the result for each gene before adjustment for multiple comparisons. Lower values indicate stronger evidence against the null hypothesis.
   • The FDR is the adjusted p-value, calculated using the Benjamini-Hochberg method, which helps control for false positives when testing many genes. Genes with an FDR below a threshold (e.g., 0.05) are considered statistically significant.

Plot Interpretations:

• MDS Plot: Displays relationships between samples based on gene expression profiles. Samples that cluster closely are more similar in their expression. Use this to identify whether samples separate by biological condition or to detect potential batch effects.
• BCV Plot: Shows the dispersion for each gene, with higher values indicating greater variability. This is useful for assessing how variability is modeled in the dataset.
• QL Plot: Highlights the quasi-likelihood dispersions, which represent variability modeled during statistical testing. Proper dispersion modeling ensures robust differential expression analysis.
• MD Plot: Visualizes the mean expression levels against log fold change for each gene. Genes far from the center indicate stronger differential expression, with points above or below the horizontal line showing upregulated or downregulated genes, respectively.

Sanitation Steps - Part 2

After performing the differential expression analysis with edgeR, we will clean the data to prepare it for visualization. This involves extracting collection elements, removing unnecessary columns, standardizing text, and splitting the file if needed. We will use the Extract element identifiers, Remove columns.

Hands On

1. Use the Extract element identifiers ( Galaxy version 0.0.2) tool with the following parameters:
   • param-file “Dataset collection”: outTables (output from the edgeR tool tool)

   Extract element identifiers will allow us to processes the edgeR output, which is a collection of datasets, to extract individual elements (like the first table from our collection) for further analysis.

2. Use the Remove columns ( Galaxy version 1.0) tool with the following parameters:
   • param-file “Tabular file”: outTables (output of edgeR tool)
   • In “Select Columns”:
     • param-repeat “Insert Select Columns”
       • “Header name”: {'id': 7, 'output_name': 'output'}
     • param-repeat “Insert Select Columns”
       • “Header name”: logFC
     • param-repeat “Insert Select Columns”
       • “Header name”: PValue
     • param-repeat “Insert Select Columns”
       • “Header name”: FDR
   • “Keep named columns”: Yes
   • “Rename output file to”: edgeR_DEG_summary

   Remove columns to filter out unnecessary columns from the edgeR output and retain only the essential ones for analysis: Gene ID (id), Log Fold Change (logFC), P-value (PValue), and False Discovery Rate (FDR).

Volcano Plot

In this step, we will use the sanitized output from the previous steps to generate a Volcano Plot, which visualizes the relationship between statistical significance (P-value) and fold change (LogFC) for differentially expressed genes (DEGs). The input file for the Volcano Plot must include four essential columns: FDR (adjusted P-value), P-value (raw), Log Fold Change, and Gene Symbols (Labels). As long as these columns are present, the Volcano Plot can be generated successfully.

Hands On: Create a Volcano Plot of the DEG

1. Use the Volcano Plot ( Galaxy version 0.0.6) tool with the following parameters:
   • param-file “Specify an input file”: edgeR_DEG_summary (tabular file output from Remove Columns tool)
   • “File has header?”: Yes
   • “FDR (adjusted P value) column number”: c4
   • “P value (raw) column number”: c3
   • “Log Fold Change column number”: c2
   • “Labels column number”: c1
   • “LogFC threshold to colour”: 0.58
   • “Points to label”: Significant
     • “Only label top most significant”: 40
   • In “Plot Options”:
     • “Plot title”: Differential Expression Volcano Plot

   Tip: What does the Volcano Plot show?

   The Volcano Plot highlights genes with statistical significance (low P-values) and large fold changes. Genes meeting the significance thresholds are typically colored and labeled for easier identification.

What are the results of the Volcano Plot?

Let’s take a moment to interpret the Volcano Plot:

Question

1. What is the significance of genes located at the extremes of the plot (e.g., high LogFC and low P-value)?
2. How many genes meet the significance threshold in this analysis?

Solution

1. Genes at the extremes of the Volcano plot are highly significant and exhibit strong differential expression. Consider exploring the biological significance of these genes to gain deeper insights into their roles.
2. In our current analysis, we identified only one downregulated gene, MTND1P23, with a logFC of -3.1956, in normal conditions compared to COVID-19 patients.

Subsetting Samples from the Original AnnData Object

In our previous analysis, we found that only one gene, MTND1P23, was identified as downregulated in our contrast. This result was obtained by “bulking” all cell types from our dataset and performing a differential expression analysis comparing normal vs. COVID-19 samples.

Now, what if we refine our approach? For instance, instead of analysing all cell types together, what happens if we focus on a specific cluster, such as T cells, and perform the same comparison: normal vs. COVID-19? Would the results remain the same, or would this approach reveal additional insights?

Extracting observations of interest as AnnData object

Hands On: Use Manipulate AnnData Tools to extract observations

1. Use the Scanpy filter ( Galaxy version 1.10.2+galaxy3) tool with the following parameters:
   • “Annotated data matrix”: AnnData for Pseudobulk (Your preprocessed, analysed, and annotated AnnData object)
   • “Method used for filtering”: Filter on any column of observations or variables
   • “What to filter?”: Observations (obs) (For filtering cells, select Observations (obs))
   • “Type of filtering?”: By key (column) values
   • “Key to filter”: cell_type (the label that identifies cell annotations in our AnnData)
   • “Type of value to filter”: Text
   • “Filter”: equal to
   • “Value”: T cell (the name of the cluster of interest for subset analysis)

After using the Scanpy filter ( Galaxy version 1.10.2+galaxy3) tool to subset the cell type of interest, go back to the top of this tutorial to the hands-on Pseudobulk with Decoupler step, and you may perform once again the same steps in this smaller AnnData object that now should only include your T cells. Results from this analysis will correspond to differential expression between conditions only for T cells.

Question

1. What data is included in the new pseudobulk count matrix. How is the matrix structured, and what do the column labels represent?
2. How many samples are included in the current dataset? Are all of them derived exclusively from T cells?
3. After performing pseudobulk analysis of T cells only, how does the volcano plot look like? Does it show differentially expressed genes between diseases?

Solution

1. The new count matrix consists of the original 2,815 rows, representing all genes with their gene labels in the first column. Additionally, it includes seven other columns corresponding to individual samples, like Control#1Tcell or SARSCoV2posSevere#1Tcell.
2. Our dataset now includes a total of seven samples: three controls and four COVID-19 positive samples.
3. The volcano plot for T cells, comparing differentially expressed genes between conditions, is shown here: No differentially expressed genes (DEGs) were detected when analyzing T cells alone between healthy and COVID-19 conditions.

Recommendations

1. Data Preprocessing:
   • Ensure raw counts are preserved in your AnnData object before pseudobulk analysis.
   • Verify the presence of important metadata fields to ensure meaningful grouping and contrasts.
2. Focus on Subsets:
   • Analyse subsets of cell types or clusters for more granular insights. For example, pseudobulk analysis focused on specific immune cell populations can reveal subtle but important expression changes.
3. Quality Control:
   • Always inspect the outputs of Decoupler (e.g., pseudobulk plots and filtering summaries) to confirm data integrity and address outliers.
   • Use edgeR’s diagnostic plots (e.g., MDS, BCV) to check for batch effects or unexpected variability.
4. Follow-Up Analysis:
   • Consider functional analysis (e.g., GO enrichment, pathway analysis) for the differentially expressed genes to link expression changes to biological processes.
   • Use visualization tools, such as heatmaps or boxplots, to further explore expression trends in key genes.
5. Future Applications:
   • Pseudobulk analysis can be extended to multi-condition experiments, time-course data, or other cell-type-specific investigations.

You've Finished the Tutorial

Please also consider filling out the Feedback Form as well!

Key points

• The advantage of pseudobulk analysis is that it bridges single-cell and bulk RNA-seq approaches, combining high resolution with statistical robustness.

• Metadata is important because proper annotation and metadata, such as cell types and conditions, are crucial for generating pseudobulk matrices.

• Decoupler plays a key role by generating pseudobulk matrices with flexibility in filtering and visualization.

• edgeR provides robust differential expression analysis, taking into account biological and technical variability.

• Visualization and interpretation are essential, with tools like Volcano Plots highlighting significant genes and trends in differential expression.

• Before starting this tutorial, prior knowledge of PBMC analysis and combining single-cell datasets is recommended.

Frequently Asked Questions

• Tutorial FAQs
• Topic FAQs
• Galaxy FAQs
• GTN Matrix Chat
• Galaxy Help Forum

Useful literature

Further information, including links to documentation and original publications, regarding the tools, analysis techniques and the interpretation of results described in this tutorial can be found here.

References

1. Liu, R., A. Z. Holik, S. Su, N. Jansz, K. Chen et al., 2015 Why weight? Modelling sample and observational level variability improves power in RNA-seq analyses. Nucleic Acids Research 43: e97. 10.1093/nar/gkv412
2. Chen, Y., and G. K. Smyth, 2016 It’s DE-licious: a recipe for differential expression analyses of RNA-seq experiments using quasi-likelihood methods in edgeR, pp. 391–416 in Methods in Molecular Biology, Springer. This book chapter explains the glmQLFit and glmQLFTest functions, which are alternatives to glmFit and glmLRT. They replace the chi-square approximation to the likelihood ratio statistic with a quasi-likelihood F-test, resulting in more conservative and rigorous type I error rate control.. 10.1007/978-1-4939-3578-9_19
3. Silvin, A., N. Chapuis, G. Dunsmore, A. G. Goubet, A. Dubuisson et al., 2020 Elevated Calprotectin and Abnormal Myeloid Cell Subsets Discriminate Severe from Mild COVID-19. Cell 182: 1401–1418.e18. 10.1016/j.cell.2020.08.002
4. Squair, J. W., M. Gautier, C. Kathe, M. A. Anderson, N. D. James et al., 2021 Confronting false discoveries in single-cell differential expression. Nature Communications 12: 5692. 10.1038/s41467-021-25960-2
5. Mompel, P. B.-i, J. V. Santiago, J. Braunger, C. Geiss, D. Dimitrov et al., 2022 decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinformatics Advances 2: 10.1093/bioadv/vbac016
6. Murphy, A. E., and N. G. Skene, 2022 A balanced measure shows superior performance of pseudobulk methods in single-cell RNA-sequencing analysis. Nature Communications 13: 7851. 10.1038/s41467-022-35519-4
7. Chen, Y., D. McCarthy, P. Baldoni, M. Ritchie, M. Robinson et al., 2024 edgeR User’s Guide. Bioconductor. First edition 17 September 2008. https://www.bioconductor.org/packages/devel/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf
8. Lab, S., and Contributors Pseudobulk Analysis Notebook: A tutorial for pseudobulk analysis using Decoupler. Accessed via Decoupler documentation: \urlhttps://decoupler-py.readthedocs.io/en/latest/notebooks/pseudobulk.html. https://github.com/saezlab/decoupler-py/blob/main/docs/source/notebooks/pseudobulk.ipynb

Feedback

Did you use this material as an instructor? Feel free to give us feedback on how it went.
Did you use this material as a learner or student? Click the form below to leave feedback.

Citing this Tutorial

1. Diana Chiang Jurado, Pseudobulk Analysis with Decoupler and EdgeR (Galaxy Training Materials). https://training.galaxyproject.org/training-material/topics/single-cell/tutorials/pseudobulk-analysis/tutorial.html Online; accessed TODAY
2. Hiltemann, Saskia, Rasche, Helena et al., 2023 Galaxy Training: A Powerful Framework for Teaching! PLOS Computational Biology 10.1371/journal.pcbi.1010752
3. Batut et al., 2018 Community-Driven Data Analysis Training for Biology Cell Systems 10.1016/j.cels.2018.05.012

BibTeX

@misc{single-cell-pseudobulk-analysis,
author = "Diana Chiang Jurado",
	title = "Pseudobulk Analysis with Decoupler and EdgeR (Galaxy Training Materials)",
	year = "",
	month = "",
	day = "",
	url = "\url{https://training.galaxyproject.org/training-material/topics/single-cell/tutorials/pseudobulk-analysis/tutorial.html}",
	note = "[Online; accessed TODAY]"
}
@article{Hiltemann_2023,
	doi = {10.1371/journal.pcbi.1010752},
	url = {https://doi.org/10.1371%2Fjournal.pcbi.1010752},
	year = 2023,
	month = {jan},
	publisher = {Public Library of Science ({PLoS})},
	volume = {19},
	number = {1},
	pages = {e1010752},
	author = {Saskia Hiltemann and Helena Rasche and Simon Gladman and Hans-Rudolf Hotz and Delphine Larivi{\`{e}}re and Daniel Blankenberg and Pratik D. Jagtap and Thomas Wollmann and Anthony Bretaudeau and Nadia Gou{\'{e}} and Timothy J. Griffin and Coline Royaux and Yvan Le Bras and Subina Mehta and Anna Syme and Frederik Coppens and Bert Droesbeke and Nicola Soranzo and Wendi Bacon and Fotis Psomopoulos and Crist{\'{o}}bal Gallardo-Alba and John Davis and Melanie Christine Föll and Matthias Fahrner and Maria A. Doyle and Beatriz Serrano-Solano and Anne Claire Fouilloux and Peter van Heusden and Wolfgang Maier and Dave Clements and Florian Heyl and Björn Grüning and B{\'{e}}r{\'{e}}nice Batut and},
	editor = {Francis Ouellette},
	title = {Galaxy Training: A powerful framework for teaching!},
	journal = {PLoS Comput Biol}
}

                   

Do you want to extend your knowledge?

Follow one of our recommended follow-up trainings:

• tutorial Hands-on: Importing files from public atlases
• slides Slides: GO Enrichment Analysis on Single-Cell RNA-Seq Data
• tutorial Hands-on: GO Enrichment Analysis on Single-Cell RNA-Seq Data

Galaxy Administrators: Install the missing tools

You can use Ephemeris's shed-tools install command to install the tools used in this tutorial.

shed-tools install [-g GALAXY] [-a API_KEY] -t <(curl https://training.galaxyproject.org/training-material/api/topics/single-cell/tutorials/pseudobulk-analysis/tutorial.json | jq .admin_install_yaml -r)

Alternatively you can copy and paste the following YAML

---
install_tool_dependencies: true
install_repository_dependencies: true
install_resolver_dependencies: true
tools:
- name: split_file_to_collection
  owner: bgruening
  revisions: 2dae863c8f42
  tool_panel_section_label: Collection Operations
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: text_processing
  owner: bgruening
  revisions: 86755160afbf
  tool_panel_section_label: Text Manipulation
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: text_processing
  owner: bgruening
  revisions: 86755160afbf
  tool_panel_section_label: Text Manipulation
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: text_processing
  owner: bgruening
  revisions: 86755160afbf
  tool_panel_section_label: Text Manipulation
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: decoupler_pseudobulk
  owner: ebi-gxa
  revisions: '09c833d9b03b'
  tool_panel_section_label: Single-cell
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: collection_element_identifiers
  owner: iuc
  revisions: d3c07d270a50
  tool_panel_section_label: Collection Operations
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: column_remove_by_header
  owner: iuc
  revisions: 2040e4c2750a
  tool_panel_section_label: Text Manipulation
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: edger
  owner: iuc
  revisions: ae2aad0a6d50
  tool_panel_section_label: RNA Analysis
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: scanpy_filter
  owner: iuc
  revisions: 5b3c1679d29b
  tool_panel_section_label: Single-cell
  tool_shed_url: https://toolshed.g2.bx.psu.edu/
- name: volcanoplot
  owner: iuc
  revisions: 99ace6c1ff57
  tool_panel_section_label: Graph/Display Data
  tool_shed_url: https://toolshed.g2.bx.psu.edu/

Feedback

No feedback has been recieved yet for this training. Be the first one by filling in the feedback form.