R Code Module 3

SummarizedExperiment

An R object that combines assay data (the count matrix), ColData (sample metadata), and RowData (gene metadata)

Assay data

The count matrix representing the read counts for each gene in each sample

ColData

The sample metadata, also known as phenotype data

RowData

The gene metadata, also known as feature data

Count

based data

Over

dispersed data

Zero

inflated data

Library size dependent data

Data where the total counts vary between samples

DESeq2

The most popular package for differential expression analysis that models count data using negative binomial distributions, accounts for over

edgeR

An alternative R package approach for differential expression

limma

voom

sleuth

A tool used for transcript

airway

An R package containing an example RNA

ggplot2

An R package that creates publication

dplyr

An R package that provides tools for data manipulation and transformation

set.seed()

A function that ensures reproducible results by fixing the random number generator

data()

A function used to load datasets, such as the airway dataset, into R's workspace

class()

A function that shows what type of R object a variable is

dim()

A function that shows the dimensions of a dataset, such as the number of genes (rows) and samples (columns)

colData()

A function that accesses or extracts the sample metadata from a SummarizedExperiment object

str()

A function that shows the structure of an object, including column names and data types

colnames()

A function that gets the sample names or column names of the count matrix

table()

A function used to count occurrences, such as how many samples are in each treatment group

rowData()

A function that extracts gene metadata from the SummarizedExperiment object

assay()

A function that accesses the actual count matrix, which is the core RNA

head()

A function that shows the first few rows of data to help understand the available information

summary()

A function that provides a statistical summary of data, such as counts

is.na()

A function that checks for missing values in the count matrix

colSums()

A function that calculates library sizes by finding the total counts per sample

rowSums()

A function that calculates the total expression per gene across all samples

colMeans()

A function that calculates the average expression per sample to help identify samples with unusual expression patterns

data.frame()

A function that creates a data frame combining variables like library sizes and average expression with treatment information for plotting

ggplot()

A function that creates plots to visualize relationships between data points

scale_y_log10()

A ggplot function that uses a log scale for the y

read.csv()

A function used to load data from a CSV file

gsub()

A function used to clean strings, such as removing gene IDs after a pipe symbol

strsplit()

A function used to parse strings, such as splitting TCGA sample IDs to extract project and participant informatio

Library sizes

Total read counts per sample

Gene detection

Number of expressed genes per sample

Distribution of counts

How reads are distributed across genes

Sample correlation

Similarity between samples

airway

Contains the example RNA

SummarizedExperiment

Provides the data structure for storing RNA

ggplot2

Creates publication

dplyr

Provides tools for data manipulation and transformation

Base R heatmap()

Creates heatmaps for sample correlation and other QC visualizations

set.seed(123)

Ensures reproducible results for any random sampling

colSums(count_data)

Calculates library sizes (total read counts per sample) a key QC metric

summary(library_sizes)

Provides statistical summary (min, max, mean, median) of library sizes

range(library_sizes)

Shows the range of library sizes to identify extreme differences

sd(library_sizes)

Calculates standard deviation to measure variability between samples

data.frame()

Creates a data frame combining library sizes with treatment information for plotting

ggplot()

Creates boxplots to visualize library size distributions by treatment group and creates plots comparing raw vs normalized library sizes

log(count_data + 1)

Log

rowMeans(log_counts)

Calculates geometric mean of log

apply() with median calculation

Computes size factors using the DESeq2 method (median ratio method)

exp(size_factors)

Converts log

sweep(count_data, 2, size_factors, "/")

Applies normalization by dividing each sample by its size factor

cor()

Used to calculate correlations and assess similarity between samples

Principal Component Analysis (PCA)

A dimensionality reduction technique that reduces complexity by finding the main sources of variation, visualizes relationships between samples, identifies patterns in high

airway

Contains the example RNA

SummarizedExperiment

Provides the data structure for storing RNA

ggplot2

Creates publication

dplyr

Provides tools for data manipulation and transformation

ggrepel

Adds non

set.seed(123)

Ensures reproducible results for any random sampling

log2(normalized counts + 1)

Applies log2 transformation to reduce the influence of highly expressed genes

apply(log_normalized_counts, 1, var)

Calculates variance for each gene across samples to identify most variable genes

order(gene variances, decreasing = TRUE)

Ranks genes by variance (most variable first)

top_variable genes <

...

t(log_normalized_counts[top_variable_genes, ])

Transposes the data matrix so samples are rows and genes are columns (required for PCA)

prcomp(pca data, scale. = TRUE)

Performs PCA with scaling to give equal weight to all genes

Experimental factors

Treatment conditions, time points, dose levels

Technical factors

Batch effects, sequencing runs, sample processing dates

Biological factors

Individual differences, cell type heterogeneity, disease subtypes

Quality issues

Sample degradation, contamination, technical failures

Real

world data

Patient heterogeneity

Each sample represents a different individual with unique genetic backgrounds, ages, lifestyles, and disease progression

Technical variation

Samples were processed at different times, in different laboratories, using potentially different protocols

Biological complexity

Cancer is inherently heterogeneous, with multiple molecular subtypes and varying degrees of disease severity

Controlled experiment

The airway dataset comes from a carefully designed laboratory experiment

Minimal biological variation

Samples are from the same cell line, treated under controlled conditions

Standardized protocols

All samples were processed using identical methods and timing

Clear treatment groups

The experimental design creates distinct, well