Home   Research   Resource   About us   Sitemap   
R Packages
 Funmap Package
 EIP Package
 R/fGWAS2 Package
 HaruPDF Package
Software
 3FunMap Software
 fGWAS2 Software

Home>>Resource >> Software

R/fGWAS2: Functional GWAS package for R (Version 2.0)



Description:

The R/fGWAS2 (Functional Genome-wide Association Studies) is developed as a new package for genome-wide association studies based on a single SNP analysis . It provides three separate methods.

Dynamic model(original fGWAS model)[1],[2]
This is a variable selection method based on FunMap model for high dimensional SNP data and longitudinal phenotype data. The model finds significant SNPs and estimates independent additive and dominant effect for each SNP. Nonparametric curve is used to structure mean vector for longitudinal data and FunMap method is employed to estimate LR2 for each SNP. The significant SNPs selected from high dimensional SNP can be outputted to the end-user or used in further process, such as Group Bayesian Lasso model.

Bayesian Lasso[3]
This is a data analysis for high dimensional SNP data and nonlongitudinal phenotype data, The whole analysis includes two steps which is characterized by the production of a preconditioned reponse variable by a supervised principle component analysis and the formulation of Bayesian lasso for selecting a subset of significant SNPs. Based on this theoretical model, the package integrates two steps into one function to analyse the genotypical data and phenotypic data. The analysis flow can be started by the data loading, and then the data goes through the main function of the Bayesian Lasso model. Finally the results will be generated. By the result summarization, the information about significant SNPs positions can be reported.

Group Bayesian Lasso[4]
This is a variable selection method based on Group Lasso model for high dimensional SNP data and longitudinal phenotype data. This model aims to select the significant SNP and estimate joint effects over the selected significant SNPs. Nonparametric curve is used to structure mean vector and MCMC sampling is employed to estimate joint additive and dominant effect for the significant SNPs.

[1]: Jiangtao Luo, Arthur Berg, Kwangi Ahn, Kiranmoy Das, Jiahan Li, Zhong Wang, Yao Li, Rongling Wu. Functional genome-wide association studies of longitudinal traits. In: Handbook of Adaptive Designs in Pharmaceutical and Clinical Development 2010, edited by S. C. Chow. Wiley, London, UK.

[2]: Kiranmoy Das, Jiahan Li, Zhong Wang, Chunfa Tong, Guifang Fu, Yao Li, Meng Xu, Kwangmi Ahn, David Mauger and Runze Li, et al. A dynamic model for genome-wide association studies . Human Genetics(2011) doi: 10.1007/s00439-011-0960-6

[3]: Jiahan Li, Kiranmoy Das, Guifang fu, Runze Li and Rongling Wu. The Bayesian Lasso for Genome-wide Associations Studies. Bioinformatics (2011) 27 (4): 516-523. doi: 10.1093/bioinformatics/btq688

[4]: Jiahan Li, Runze Li, and Rongling Wu. Bayesian Group LASSO for Varying-Coefficient Models With Application to Functional Genome-Wide Association Studies.

Download:

The packages for Windows, Linux and Mac OS are available. The document also can be downloaded by the following link.


Installation:

The R/fGWAS2 package depends on 3 packages, including mvtnorm and MSBVAR. Before R/fGWAS2 is installed, these packages should be installed in advance. After you download this package file, please type the following command or click the menu item "Install packages from local zip files".

Windows OS:

>install.packages("x:/fullpath/fgwas_2.0.zip", repos=NULL)

Linux/Mac OS:

>install.packages("/fullpath/fgwas_2.0.tar.gz", repos=NULL)

Before the package is used in R, It is necessary to loading package by the following command:

> library(fgwas2)

After it is loaded, all functions within R/fGWAS2 will be readily available to the user.

Sample Script:

The following source shows how to call the main function in R.

Bayesian Lasso is used to analyze nonlongitudinal phenotype data with high dimensional SNP data.

> BLSApp.onestep("bmi10.tped", "bmi10.tfam", "bmi_phenos.csv")

Dynamic model is used to analyze longitudinal phenotype data with high dimensional SNP data.

> DYNApp.onestep("bmi10.tped", "bmi10.tfam", "bmi_phenos2.csv");

Group Bayesian Lasso is used to analyze longitudinal phenotype data with high dimensional SNP data.

> GLSApp.onestep("bmi10.tped", "bmi10.tfam", "bmi_phenos2.csv");


Simulation:

The following scripts show how to do a simulation of Dynamic model, Bayesian Lasso and Group Lasso model.

> DYNApp.simu_test();
> BLSApp.simu_test();
> GLSApp.simu_test();

Sample Figures:



Sample Reports:

Report of real data by Dynamic model(fGWAS).

Last updated:

Version 2.0 10/05/2012
Version 1.0 01/19/2011


Copyright© 2009-, Center for Statistical Genetics, Penn State University, All Rights Reserved.