PUMAS

Fine-tuning polygenic risk score models using GWAS summary statistics

Updates

Please clone the repo to local in order to build and load.

Last update: 10/25/2019. Please update the package if downloaded before 10/26/2019 4:53pm CT.

Introduction

Our project gives user-friendly function that presents a direct and explicit result of fine-tuning polygenic risk score models. A quick start panel that walks through PUMAS usage is included below. For full explanation and tutorial, please go to wiki

Step 0

Downloading pumas repo form this page by clicking Clone or Download

Step 1

Load our package by double click on pumas.Rproj and go to menu of R > Build > Load All

Step 2

Giving input information to do an analysis from our package in R console. In our example, see details in wiki

pumas.main(input_path="/working-directory-to-input/T0030_pruned.txt",output_path="/working-directory-to-output/T0030_pruned.png",beta_header="Beta",af_header="EAF",se_header="SE",pvalue_header="Pval",samplesize_header=766345,make_plot=TRUE)

Quick Start

Note: before applying PUMAS to GWAS summary statistics, the GWAS data needs to be pruned in advance. Do not use clumped GWAS as input.

LD-pruning can be done by PLINK.

Input Data

The function requires an input GWAS summary statistics file in the form of .txt/.txt.gz. The GWAS summary data should include:

Parameter	Example	Description
input_path	"input/T0030_pruned.txt"	GWAS path
beta_header	Beta	Effect size
af_header	EAF	Either allele/minor allele frequency
se_header	SE	Standard error
pvalue_header	Pval	P-value
samplesize_header	766345	Sample size

Optional Features

samplesize_header can be either a number or a character of column name. If the input GWAS does not include a column for per-SNP sample sizes, the user can provide a single number (usually reported along with a published GWAS) as the uniform sample size for all SNPs.

n_fold is the number of subsets that PUMAS partitions the complete GWAS into. For example, when n_fold=4 PUMAS will generate the training summary statistics based on 3 subsets and calculate the testing summary statistics of the remaining 1 subset. n_fold can be user-specified. When n_fold is not specified, PUMAS will use a default of 4 subsets to implement the model tuning approach.

odds_ratio is a boolean value that tells PUMAS whether the weight input from GWAS is effect size (quantitative) or odds ratio (binary). When odds_ratio=T, PUMAS asusmes that beta_header is the header for odds ratio and applies log-transformation on the column. The default is odds_ratio=F.

Make Plot (optional)

PUMAS can output a scatterplot that illustrates a detailed pattern of the predictive performance under each PRS models. Y-axis is the predictive R2 and X-axis is log-transformed p-value cutoff for every model. To make this plot, the user needs to provide 2 more parameters to the main function:

output_path is the path to store the plot in .png format. For example, output_path='result/test.png' is acceptable.

make_plot is a boolean value that tells PUMAS whether to output a scatterplot. The default is make_plot=F.

For more detailed interpretations please see details in wiki page.

Result

PUMAS results are printed in the interface.

Maximal.R2is the estimated R2 at for the best model selected.

Optimal.P.value is the p-value cutoff for the best model selected.

Authors

See also the list of contributors who participated in this project.

License

All rights reserved for Lu-Laboratory

Citation

If you use the package, please cite

Zhao, Z. et al. Fine-tuning Polygenic Risk Scores with GWAS Summary Statistics. bioRxiv, 810713 (2019).

Acknowledgments