comorbidPGS

comorbidPGS is a tool for analysing an already computed Polygenic Score (PGS, also named PRS/GRS for binary outcomes) distribution to investigate shared genetic aetiology in multiple conditions.

comorbidPGS is under GPL-3 license, and is freely available for download.

Prerequisite

• R version 3.5 or higher with the following packages:
  • stats
  • utils
  • ggplot2

Installation

comorbidPGS is available on CRAN, you can download it using the following command:

install.packages("comorbidPGS")

If you prefer the latest stable development version, you can download it from GitHub with:

if (!require("devtools", quietly = TRUE)) install.packages("devtools")
devtools::install_github("VP-biostat/comorbidPGS")

Example

Building an Association Table of PGS

This is a basic example which shows you how to do basic association with the example dataset:

library(comorbidPGS)
#> 
#> Attaching package: 'comorbidPGS'
#> The following object is masked from 'package:graphics':
#> 
#>     assocplot

# use the demo dataset
dataset <- comorbidData
# NOTE: The dataset must have at least 3 different columns:
# - an ID column (the first one)
# - a PGS column (must be numeric, by default it is the column named "SCORESUM" or the second column if "SCORESUM" is not present)
# - a Phenotype column, can be factors, numbers or characters

# do an association of one PGS with one Phenotype
result_1 <- assoc(dataset, prs_col = "t2d_PGS", phenotype_col = "t2d")

# do multiple associations
assoc <- expand.grid(c("t2d_PGS", "ldl_PGS"), c("ethnicity","brc","t2d","log_ldl","sbp_cat"))
result_2 <- multiassoc(df = dataset, assoc_table = assoc, covar = c("age", "sex", "gen_array"))
#> Warning in phenotype_type(df = df, phenotype_col = phenotype_col): Phenotype
#> column log_ldl is continuous and not normal, please normalise prior association

#> Warning in phenotype_type(df = df, phenotype_col = phenotype_col): Phenotype
#> column log_ldl is continuous and not normal, please normalise prior association

Extension of association analysis: one-sample MR using the Wald Ratio and 2SLS methods

# MR using Wald ratio method
mr1 <- mr_ratio(df = dataset, prs_col = "ldl_PGS", exposure_col = "log_ldl", outcome_col = "sbp") 
#> Warning in phenotype_type(df = df, phenotype_col = exposure_col): Phenotype
#> column log_ldl is continuous and not normal, please normalise prior association
#> Warning in phenotype_type(df = df, phenotype_col = outcome_col): Phenotype
#> column sbp is continuous and not normal, please normalise prior association

# MR using 2-stage least square method (2SLS)
mr2 <- mr_2sls(df = dataset, prs_col = "ldl_PGS", exposure_col = "log_ldl", outcome_col = "sbp") 
#> Warning in phenotype_type(df = df, phenotype_col = exposure_col): Phenotype
#> column log_ldl is continuous and not normal, please normalise prior association
#> Warning in phenotype_type(df = df, phenotype_col = outcome_col): Phenotype
#> column sbp is continuous and not normal, please normalise prior association

Examples of data visualisation using comorbidPGS

densityplot(dataset, prs_col = "ldl_PGS", phenotype_col = "sbp_cat")

# show multiple associations in a plot
assoplot <- assocplot(score_table = result_2)
assoplot$continuous_phenotype

assoplot$discrete_phenotype

NOTE: The score_table should have the assoc() output format

centileplot(dataset, prs_col = "brc_PGS", phenotype_col = "brc")
#> Warning in centileplot(dataset, prs_col = "brc_PGS", phenotype_col = "brc"):
#> The dataset has less than 10,000 individuals, centiles plot may not look good!
#> Use the argument decile = TRUE to adapt to small datasets

As those graphical functions use ggplot2, you can fully customize your plot:

library(ggplot2)
centileplot(dataset, prs_col = "t2d_PGS", phenotype_col = "t2d") + 
  scale_color_gradient(low = "green", high = "red")

decileboxplot(dataset, prs_col = "ldl_PGS", phenotype_col = "ldl")

Citation

To cite comorbidPGS in publications, please use: