This repository walks you through a standard approach to case/control genome wide association studies (GWAS) using the data from a Late Onset Alzheimer's Disease study (Webster et al., 2009).
This Youtube video shows all the steps for the GWAS demo.
Genome-wide association study (GWAS) is a hypotheses-free method for identifying associations between genetic regions and traits (incl. diseases). GWAS typically searches for small variations, known as single-nucleotide polymorphisms (SNPs), that occur more frequently in people with a particular disease than in people without the disease. GWAS already has seen success at identifying SNPs related to conditions such as diabetes, Parkinson's and Crohn's.
Researchers use two groups of participants: people with the disease being studied (case group) and similar people without the disease (control group). DNA is obtained from each participant by scanning their blood sample/cell on automated laboratory machines. The machines survey each participant’s genome for strategically selected markers of genetic variation (SNPs). If certain variations are found to be significantly more frequent in people with the disease compared to people without the disease, the variations are said to be associated with the disease.
The SNPs associated with the disease are identified by testing for statistical significance between cases and controls. Results are typically displayed in a Manhattan plot with -log10(p-value) plotted against the position in the genome. Two lines are added to indicate the genome-wide significance threshold (p=5.0×10−8) and the cut-off level for selecting single-nucleotide polymorphisms for replication study (p=1.0×10−5). An example Manhattan plot is illustrated below.
Webster et al. (2009) study investigated the genetic control of human brain transcript expression in Late Onset Alzheimer's Disease. The paper did multiple types of analysis but we will only implement a simple case/control GWAS using the data from this study to identify genetic variants assoaciated with the Late-Onset Alzheimer's Disease.
GWAS study included 364 participants of European descent with either confirmed LOAD or no neuropathology present (176 cases + 188 controls).
First, you will need do download the raw data, which is in bioinformatics software PLINK text format (detailed instructions can be found here). The data is composed of two files: *.map file, which contains SNP information such as chromosome number, variant ID, and so on; *.ped file contains sample pedigree information and genotype calls.
The GWAS analysis was divided into three steps:
- Data Quality Control (QC): remove samples and variants that do not meet the quality control criteria.
- Association testing: statistical significance testing between cases and controls.
- Results visualisation. The results are typically displayed in a Manhattan plot.
For the QC step, two Jupyter Notebooks were produced: a general and a detailed version. The latter one walks you throught the QC process step-by-step explaining all the underlying concepts.
To run these Notebooks, you will need to have PLINK 1.90 beta installed as a command line tool. The installation instructions can be found here. The Notebook is written in R, so you also need an R kernel, which can be easily downloaded by opening an R console and running:
install.packages('IRkernel')
Then you also need to make the kernel available to Jupyter. The following command will install the kernel spec system-wide. If you want it to be available just to the current user, set user to TRUE:
IRkernel::installspec(user = FALSE)
Running cells in the notebook executes PLINK commands in the background. If you want to see what each step did, you can check the *.log files in the output folder.
Once the QC step is complete, association testing is also done from the same notebook by calling PLINK.
A Shiny app was created for visualizing the GWAS results.
The app can be run from the R console:
library(shiny)
runApp("app.R")
The app has four tabs:
- An interactive Manhattan plot showing the the negative logarithm of the p-value plotted against the position in the genome. The red line represents a genome-wide significance threshold, thus SNPs above it will be associated with the Late Onset Alzheimer's Disease. You can hover your mouse over the SNPs to display the variant information.
- Circular and Rectangular Manhattan plots: not interactive, more controls for changing the plot parameters as well as functionality for zooming into a single chromosome.
- Quantile-quantile (QQ) plots, showing the observed vs expected p value.
- SNP density plots.
