Independent Study to prepare for workshop

Data Analysis 2: Immunobiology - Sample data analysis

Emma Rand

Overview

There are three workshops:

  1. Week 2 DA 2: Immunobiology - Sample data analysis

  2. Week 4 DA 3: Immunobiology - Analysis of your own data

  3. Week 6 DA 4: Immunobiology - Customising figures

Overview

These slides:

  • Prepare you for the sample data analysis workshop; your own data will be like the sample data

  • Summarise the experimental design and aims

  • Explain what the data are

  • Go through the analytical steps conceptually

  • Explain what tools we will use in the workshop to do the analysis

Experimental design and aims

Experimental design and aims

  • Macrophages produce TNF-α in response to bacterial infection

  • Question: Does the production of TNF-α by macrophages require live bacteria, or is the cell wall component sufficient?

  • Therefore we need 3 treatments: Media (control), Lipopolysaccharide (LPS, Cell wall component of E. coli) and live E. coli

  • We measure TNF-α with a TNF-α antibody conjugated to Allophycocyanin (APC)

  • Therefore we need a control for antibody binding and use Isotype antibody

Experimental design and aims

  • Macrophages are treated with one of three treatments: Media, LPS E. coli which are NeonGreen fluorescence

  • Two antibodies are used for each treatment: Isotype antibody, TNF-α antibody conjugated to Allophycocyanin (APC)

  • Thus there are 3 x 2 = 6 combinations

  • Two variable of interest: red fluorescence, green fluorescence

  • We also measure forward scatter (cell size) and side scatter (cell granularity) which can be used to quality control the cells

Experimental design and aims

  • We only expect to see red fluorescence (APC) if the treatment induces TNF-α production in macrophages and the TNF-α antibody is used.

  • We only expect to see green fluorescence (FITC) if the treatment is E. coli

  • This is summarised in the figure on the next page

Experimental design and aims

The data

The data

  • The data are in a flow cytometry standard format (FCS) file

  • Each FCS file contains data from one sample

  • You will have 6 FCS files, one for each combination of treatment and antibody

  • there are 22 variables in columns and up to 50000 cells in rows

The data

  • the 22 columns: TIME, Time MSW, Pulse Width, FS Lin, FS Area, FS Log, SS Lin, SS Area, SS Log, FL 1 Lin, FL 1 Area, FL 1 Log, FL 2 Lin, FL 2 Area, FL 2 Log, FL 3 Lin, FL 3 Area, FL 3 Log, FL 8 Lin, FL 8 Area, FL 8 Log, Event Count

  • FS is Forward scatter, SS is Side scatter, FL is fluorescence channel

  • FL 1 is the green fluorescence channel and we will rename it E_coli_FITC

  • FL 8 is the red fluorescence channel and we will rename it TNFa_APC

  • FL 2 and FL 3 are not used in this experiment and we will delete those columns

  • We will use just four columns: E_coli_FITC_Lin, TNFa_APC_Lin, FS Lin, and SS Lin

Analytical steps

Overview

  • The analysis of flow cytometry data is relatively simple conceptually

  • We apply several quality control steps to the data to remove anomalous signals, dead cells and debris

  • We use scatter plots, calculate means, and find percentages of cells in different regions of the scatter plots

Analytical steps

  • Import the data into R, improve the column names and remove unwanted columns

  • Apply automated quality control

  • Apply a “logicle” transformation (Parks, Roederer, and Moore 2006) to the fluorescence channels (similar to logging)

  • Explore the data with scatter plots and histograms/density plots

  • Use FS Lin and SS Lin to determine what cells (rows) to remove as debris

  • Determine cut-offs for cells being positive for TNF-α and E. coli

  • Calculate the percentage of cells that are positive for TNF-α for each treatment

Tools

Tools

  • Import, rename and subset columns using the flowCore package (Ellis et al. 2024)

  • Automated quality control with the flowAI package (Monaco et al. 2016)

  • Apply a “logicle” transformation using the flowCore package

  • Put the data into a dataframe to make it easy to use tidyverse (Wickham et al. 2019) tools like group_by(), summarise(), ggplot(), filter()

The data in R

  • the flowCore package imports each FCS file as a flowFrame object

  • The flowFrame object contains the data from the FCS file and metadata about the experiment

  • A collection of related flowFrames are stored in a flowSet object

  • flowAI and flowCore functions work with flowSet objects

  • After that we can convert the flowSet to a dataframe to use tidyverse tools

Summary

Summary

  • Sample data are like the data you will produce in your own experiment

  • 3 treatments x 2 antibodies = 6 combinations; 4 variables upto 50000 cells each

  • The analysis is conceptually simple: quality control, transformation, scatter plots, and calculating percentages

  • The week 2 workshop analyses the sample data, in the week 4 workshop you will analyse your own data

  • We will use the flowCore, flowAI and tidyverse packages to do the analysis

References

🔗 Data Analysis 2: Immunobiology - Sample data analysis

Ellis, B, Perry Haaland, Florian Hahne, Nolwenn Le Meur, Nishant Gopalakrishnan, Josef Spidlen, Mike Jiang, and Greg Finak. 2024. “flowCore: flowCore: Basic Structures for Flow Cytometry Data.”
Monaco, Gianni, Hao Chen, Michael Poidinger, Jinmiao Chen, Joao Pedro de Magalhaes, and Anis Larbi. 2016. “flowAI: Automatic and Interactive Anomaly Discerning Tools for Flow Cytometry Data” 32. 10.1093/bioinformatics/btw191.
Parks, David R., Mario Roederer, and Wayne A. Moore. 2006. “A New “Logicle” Display Method Avoids Deceptive Effects of Logarithmic Scaling for Low Signals and Compensated Data.” Cytometry Part A 69A (6): 541–51. https://doi.org/10.1002/cyto.a.20258.
Wickham, Hadley, Mara Averick, Jennifer Bryan, Winston Chang, Lucy D’Agostino McGowan, Romain François, Garrett Grolemund, et al. 2019. “Welcome to the Tidyverse” 4: 1686. https://doi.org/10.21105/joss.01686.