Analysis code for de Jong et al. 2019

This repository contains code used in the analysis of data from de Jong et al. 2019.

All of the code (and data) is deposited as a static snapshot in the Cambridge Apollo repository (https://doi.org/10.17863/CAM.20379), corresponding to release 0.1 of the GitHub repository.

The code in the GitHub repository is non-static and may be updated if we find any bugs (see "NEWS" section at the bottom).

Setup

1. Clone the repository (or download as a zip file and extract the files) - you should then have a project directory called "publication_deJong2019_Nplasticity"
2. Download and extract the zip files with the raw data and the processed data from the Appolo repository. Place them inside the previous project directory.

In the end you should have this directory structure:

publication_deJong2019_Nplasticity
          ├── data
          ├── data_processed
          └── scripts

Running analysis

Most of the analysis was done using R.

To recreate the analysis and figures for the paper, open the 2019_deJong_Nplas.Rproj RStudio project file (which will ensure you have the correct working directory set - all R scripts use relative paths from the project directory).

There is also a document detailing how the mixed models were fit to trait data

Some details about the files on scripts/R:

• All scripts named with prefix Fig/Table recreate each figure/table in the paper. Some of the graphs were slightly edited in inkscape for the published version.
• The scripts with numbered prefix 01-04 do different data processing steps and produce files already provided in the data_processed folder in our repository. However, they can be used to recreate those analysis.
  • 01 - tidy the phenotype data
  • 02 - fit linear mixed models on trait data (used to estimate variance components and heritabilities)
  • 03 - run standard QTL scans using R/qtl2 package
  • 04 - run custom QTL scans using a mixed model approach, as well as a permutation for significance thresholds.

The GWAS was done using standalone command-line programs. The code is in the scripts/shell directory. Those scripts are numbered according to the order in which analysis was run. This part of the analysis is not completely reproducible in its current form, as we used a pre-release version of the Arabidopsis genotype data. However, a similar analysis should be possible to run using the publicly available SNP data.

Finally, we used the python package LIMIX to fit a multi-SNP QTL model (scripts/python/limix_multitrait.py) as published by Sasaki et al 2015.

Software versions

# R and packages
R 3.5.3
tidyverse 1.2.1
patchwork 0.0.1
broom 0.5.2
qtl2 0.18
atMAGIC 0.1.0

# python and LIMIX
python 2.7.12
limix 0.7.12