(2023) Analysis Using R

Module 1: Exploratory data analysis Overview & Clustering

• Knowing your data: An overall workflow for exploratory data analysis
  • Understand the difference between response variables, explanatory variables, biological variation, technical variation, and batch effects
  • Missing data; understand how to identify structured versus unstructured missingness, and the role of imputation
  • Finding unwanted sources of variation; surrogate variable analysis and RUVseq

• Knowing your data’s structure. Calculating “distance” between (high-dimensional) data points
  • What distance metrics represent
  • Different kinds of different metrics and when to use them
• Clustering principles & methods
  • Why cluster?
  • A survey of clustering methods
  • Choose the clustering method that is right for your data
• Assessing the quality of clustering results
  • Metrics for identifying the optimal number of clusters
  • Existential questions introduced by clustering

Module 2: Dimensionality reduction

• What is dimensionality reduction, and common applications in bioinformatics
• Dimensionality reduction with Principal Components Analysis (PCA)
  • Conduct PCA on different types of data
  • Get information out of PCA objects in R 
• Some practical uses of PCA 
  • Plot and learn from PCA output
  • Use PCs as control variables in your analysis
  • Use PCs as variables of interest in your analysis
• Other types of dimensionality reduction
  • t-stochastic neighbor embedding (tSNE) 
  • uniform manifold approximation and project (UMAP)

Module 3: Fitting generalized linear models

• Read different data files into R

• Merge data and handle missing values
• Use ggplot to create and modify publication-quality R plots
• Plot and fit linear model for continuous-valued outcome, and logistic model for dichotomous outcome

Module 4: Differential expression analysis

• Manually conduct many parallel statistical tests
  • Different types of statistical tests
  • Evaluate and plot output 
  • Extract output for tables 
  • Visualize p-values from multiple testing: QQplot, volcano plot.
  • Correct for multiple statistical tests: Bonferroni, false discovery rate
• Using bioconductor for analysis
  • Perform differential expression analysis