Identification and analysis of differentially expressed genes in gene-expression datasets

The Gene Expression Omnibus (GEO) is a publicly-funded repository of gene-expression studies maintained by the NCBI. Access the repository here. Example GEO datasets are in the table below

Perform these tasks to familiarize yourself with respect to basis analyses of this type of data. (Note that some of the tasks can be performed using the GEO web interface - see below. However, for the purposes of the workshop perform these analyses yourself in R)

• Select a GEO Dataset, and follow the links to read about the study
• Make sure to understand the study and the experimental design
• What is the objective of the study?
• What platform has been used and what has been measured?
• What are the sample groups?
• How many replicates have been performed, and what type of replicates are they?
• Data analyses and visualizations
• (use R code for the analysis)
• Visualize the distributions of the data (for example using histograms or boxplots)
• Does the data need to be normalized or scaled?
• Visualize the correlation between samples (for example using pairwise scatterplots and computing correlations)
• Use PCA to visualize the structure in the data. Do the sample groups cluster?
• Compute the ratio (use the log2 ratio) for pairwise comparison(s) within the dataset
• Identify DEGs within the dataset (use T-test)
• Note that for the purposes of this workshop we do not do any multiple testing correction
• Plot Volcano plot(s) using the ratio and T-test p-values
• Biological interpretation
• How many DEGs are there with e.g. p-value < 0.05?
• Identify functional gene networks using STRING
• Identify enriched pathways and processes using e.g. GOrilla or Enrichr
• How does your biological interpretation relate to the original study? What could you add to the analysis?

Case Studies (GEO Datasets) from GEO

Study Title	GEO Link	Data
PTEN deletion mutation effect on colon cancer cells	GDS2446	mat.GDS2446.csv
Colorectal adenoma formation	GDS2947	mat.GDS2947.csv
Beta-catenin inactivation effect on intestinal crypts	GDS2984	mat.GDS2984.csv
Beta-catenin deficiency model: kidney	GDS3322	mat.GDS3322.csv
Colon carcinoma response to butyrate and aspirin	GDS332	mat.GDS332.csv
Beta-catenin depletion effect on pancreatic cancer cell line	GDS3578	mat.GDS3578.csv
Colorectal cancer tumors	GDS4382	mat.GDS4382.csv
Colorectal cancer tumors	GDS4379	mat.GDS4379.csv
Ls174T colon cancer cell line response to Wnt signaling inhibition	GDS4386	mat.GDS4386.csv
5-aza-2-deoxycytidine effect on colorectal cancer cell lines	GDS4397	mat.GDS4397.csv
Model of -catenin overexpression in embryonic kidney	GDS4449	mat.GDS4449.csv
Sporadic medulloblastomas	GDS4469	mat.GDS4469.csv
Medulloblastomas in children	GDS4471	mat.GDS4471.csv
MYC-driven medulloblastoma model	GDS4478	mat.GDS4478.csv
Colorectal cancer cell line SW480 response to Snail overexpression	GDS4596	mat.GDS4596.csv
Central nervous system primitive neuroectodermal tumors	GDS4838	mat.GDS4838.csv
Beta-catenin depletion effect on pancreatic cancer cell line	GDS5324	mat.GDS5324.csv

Using the GEO web interface to identify differentially expressed genes in gene-expression datasets

See the video tutorial here

• Select a data set
  • http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE44097
• Click 'Analyze with GEO2R'
  • http://www.ncbi.nlm.nih.gov/geo/geo2r/?acc=GSE44097
• Define groups of samples (e.g. treated and control)
• Assign samples to groups
• Click 'Top 250' to give you list of differentially expressed genes