Implement code to execute an X-wide association study (XWAS) in telomere length

• X-wide association study (XWAS) is a data-driven method to find what variables in the set 'X' are associated with a phenotype (call it Y, e.g., telomere length)
• Originally presented at ISEE/ISES 2018, Ottawa, Canada on August 26 2018

Contact information:

• Chirag J Patel (chirag hms dot harvard dot edu)
• Twitter/GitHub: @chiragjp (https://twitter.com/chiragjp)
• web: http://www.chiragjpgroup.org
• Funding support: NIEHS R00 ES023504, NIEHS R21 ES025052, NIAID R01 AI127250, NSF 1636870

Prerequisites for the tutorial:

• familiarity with programming and data structures
• familiarity with R and regression

Software requirements

• R (version 3.3.3 or greater)
• Rstudio (version 1.01 or greater)
• Packages: survey, tidyverse, knitr, and broom

Execution

• download 'xwas_tutorial.Rmd' in RStudio from here: https://github.com/chiragjp/xwas_with_nhanes_tutorial/blob/master/xwas_tutorial.Rmd
• knit the .Rmd in Rstudio
• or download a pre-knitted file: https://github.com/chiragjp/xwas_with_nhanes_tutorial/blob/master/xwas_tutorial.html

Slide Downloads

• Part 1, Data analytics to enable X-wide Association Studies (XWASs): https://github.com/chiragjp/xwas_with_nhanes_tutorial/blob/master/ISEE%20Data%20Analytics%20Workshop%20083018.pdf
• Part 2, Introduction to the hands-on tutorial to get you started to executing X-wide Association Studies (XWASs) with survey data: https://github.com/chiragjp/xwas_with_nhanes_tutorial/blob/master/ISEE%20Data%20Analytics%20Hands%20On%20082618.pdf

Original paper:

• Patel CJ, et al, International Journal of Epidemiology 2016
• https://www.ncbi.nlm.nih.gov/pubmed/27059547

Questions:

• Send a 'pull' request in GitHub to incorporate your answers in a new version of the .Rmd or R code!
• https://services.github.com/on-demand/github-cli/open-pull-request-github

0. What is your interpretation of the association sizes and pvalues for your XWAS?
1. How much to the coefficients change for the adjustment variables for each of the correlations?
2. Attempt to reproduce the findings in Patel et al., IJE 2016 (https://www.ncbi.nlm.nih.gov/pubmed/27059547) using more categories of exposure. How will you handle other X variables, such as self-reported variables?
3. When increasing the number of variables, how would the pvalue threshold change to accommodate more tests? How would the FDR change, if at all?
4. Execute the XWAS in another phenotype. What are the similarties and differences between your analysis in 1.
5. How much variance explained in telomeres do the top factors explain? Is this to be expected?
6. Implement the XWAS without using the for operator using the tidyverse suite of commands.