fboehm/manhattan

Introduction

The Manhattan plot is a specialized form of scatterplot to display genome-wide association studies (GWAS). The x-axis of a Manhattan plot is the genomic position, and the y-axis is usually the $-log_{10}(P\text{-}value)$ (although other sensible metric can be used as well).

There are many packages for making Manhattan plots, but most of them are not easily extensible. The package ggplot2 has become increasingly popular among the $\textbf{R}$ and bioinformatics communities. Therefore, a package that is fully compatible with ggplot2 and customizable with its geoms will facilitate the use of Manhattan plots.

Introducing manhattan, a ggplot2 based package for making Manhattan plots. It takes in a data.frame and returns a standard ggplot object, upon which the user can add other geoms. This makes it very convenient for users to build on manhattan and customize their plots.

Installation

To install manhattan, use the standard R package installation command.

# install.packages('manhattan')

If you want the latest development version, install it using devtools.

devtools::install_github("boxiangliu/manhattan")

The package has been tested on Linux and Mac OSX. It has not been tested on Windows.

Usage

Basic Manhattan plot

To illustate its usage, let us plot the coronary artery disease GWAS based on Deloukas et al(2013). The original dataset provides nominal p-values. Since we want to plot the $-log_{10}(P\text{-}value)$, let us take the logarithm.

library(manhattan)
data(cad_gwas)
cad_gwas$y=-log10(cad_gwas$pval)
head(cad_gwas)

The dataset contains five columns: chrom, pos, rsid, pval, and y. Three of them are required:

chrom
pos
y

The chrom and pos columns specify the genomic location (x-axis), and y specify the y-axis (duh!). The choice of the column name "y" is intentional - not every Manhanttan plot uses $-log_{10}{P\text{-}value}$ as the y-axis. After loading the data, we are ready to make a Manhattan plot. Notice that Deloukas et al uses hg18. To get the chromosome lengths correctly, we specify hg18 as an argument.

manhattan(cad_gwas,build='hg18')

Ta-da! Our first Manhattan plot.

Customizing colors

If black and grey are dull, we can change the color of each chromosome.

manhattan(cad_gwas,build='hg18',color1='skyblue',color2='navyblue')

Highlight and label SNPs and genes

A common task is to highlight and annotate SNPs of interest. The package manhattan requires color and SNP labels to be specified in the input data.frame as two columns: color and label. Note that only SNPs of interest have color strings, and other SNPs must be left as NA.

Let us highlight two SNPs rs602633 and rs1333045.

cad_gwas[cad_gwas$rsid=='rs602633','color']='green'
cad_gwas[cad_gwas$rsid=='rs1333045','color']='red'
manhattan(cad_gwas,build='hg18')

We could also label the two SNPs. Note again that only SNPs of interest have label strings, other SNPs should be left as NAs. Since manhattan returns a ggpplot object, we could just add a geom_text layer.

cad_gwas[cad_gwas$rsid=='rs602633','label']='rs602633'
cad_gwas[cad_gwas$rsid=='rs1333045','label']='rs1333045'
manhattan(cad_gwas,build='hg18')+geom_text(aes(label=label),hjust=-0.1)

Adding GWAS significance line

It is worth noting that any standard geoms can be used with manhattan. For instance, let's add a line indicating genome-wide significant threshold.

manhattan(cad_gwas,build='hg18')+geom_text(aes(label=label),hjust=-0.1)+geom_hline(yintercept=-log10(5e-8),color='red')

Plotting multiple GWAS studies

We can go even further to use facets with manhattan. For illustration, let us pretend that we have two GWAS studies by duplicating cad_gwas, and plot two GWAS studies on top of each other.

cad_gwas_2=rbind(cbind(cad_gwas,study='GWAS 1'),cbind(cad_gwas,study='GWAS 2'))
manhattan(cad_gwas_2,build='hg18')+facet_grid(study~.)

Details

Behind the scene, manhattan is no more than a wrapper around ggplot, with a few tricks to transform a genomic axis to a scatterplot axis. In brief, manhattan transforms the chrom:pos pairs to cumulative positions. For instance, chr2:1 would be the length of chromosome 1 plus 1, chr3:1 would be chromosome 1 plus chromosome 2 plus 1, so on and so forth. Therefore, it is important to specify the genomic build (e.g. hg19) so that manhattan can make the correct transformation. It then positions the chromosome labels on the x-axis according to these transformations.

Again, it is important to note that manhattan is no more than a wrapper around ggplot, which makes manhattan highly customizable. For instance, we can change the size of the SNPs of interest by adding a geom_point layer.

cad_gwas$size=ifelse(cad_gwas$rsid%in%c('rs602633','rs1333045'),5,2)
manhattan(cad_gwas,build='hg18')+geom_point(aes(color=color,size=size),show.legend=FALSE)

Questions and Bugs

If you have any question or want to report a bug, please open a github issue here.

Citation

If you use the manhattan package, please cite our paper: https://www.nature.com/articles/s41588-019-0404-0

Liu, Boxiang, Michael J. Gloudemans, Abhiram S. Rao, Erik Ingelsson, and Stephen B. Montgomery. "Abundant associations with gene expression complicate GWAS follow-up." Nature genetics 51, no. 5 (2019): 768-769.

fboehm / manhattan