JEPEGMIX

What's New?

[04/06/2016] A pre-compiled executable for JEPEGMIX (v0.2.0) built under Linux is available now!

Citing JEPEGMIX

JEPEGMIX: a summary statistics based tool for gene-level joint testing of functional variants, Donghyung Lee; Vernell S. Williamson; Tim B. Bigdeli; Brien P. Riley; Bradley T. Webb; Ayman H. Fanous; Kenneth S. Kendler; Vladimir I. Vladimirov; Silviu-Alin Bacanu; Bioinformatics (2016) Jan 15; 32(2):295-7. doi:10.1093/bioinformatics/btv567.

Acknowledgement

This work is supported by the National Institutes of Health with grants R25DA26119, R21MH100560, R21AA022717 and P50AA022537.

Download JEPEGMIX

The current release (Version 0.2.0) of JEPEGMIX is for a Linux user. The pre-compiled executables for other operating systems (e.g., Windows, MacOS) will be available soon. The latest source codes of JEPEGMIX are available upon request (dustorm_at_gmail_dot_com)

Download Reference Panels

Here are the 1000 Genome population abbreviations used by JEPEGMIX. AFR is an abbreviation for African; AMR for admixed American; ASN for East Asian; EUR for European.

Download SNP Annotation Data

JEPEGMIX Input File Format

When study allele frequency information is available

JEPEGMIX takes as input a plain text file with rows and columns denoting SNPs and variables, respectively. The first line of the input file should be column names/headers. Data entries on each line should be separated by white space. The file should contain eight columns: 1) rsid (SNP ID), 2) chr (chromosome number), 3) bp (base pair position), 4) a1 (reference allele), 5) a2 (alternative allele), 6) z (normally distributed GWAS/meta-analysis summary statistic, i.e. two-tailed Z-score), 7) info (imputation information) and 8) af1 (cohort reference allele frequency (RAF)). If your input data is generated from non-imputed genotype data (e.g. whole genome or exome sequencing data), you can set all imputation information scores equal to 1 (info=1). JEPEGMIX does not require the input data to be sorted in ascending order by chromosome number and base pair position or SNP ID. Here is a sample JEPEGMIX input file.

rsid        chr  bp         a1  a2  z          info    af1

rs1000109   9    117908733  C   T  -0.714464   0.890   0.384

rs10001109  4    44904653   C   T   0.721919   0.731   0.198

rs1000112   22   28273339   C   G  -0.583666   0.445   0.042 

rs10001127  4    130547376  T   C   0.069329   0.999   0.210 

rs1000113   5    150240076  T   C  -1.447288   1.0     0.189

When study allele frequency information is not available

Genome-wide association studies/meta-analyses typically do not provide study allele frequency information due to privacy issues. However, information about ethnic proportion of the study samples is usually available from the publications. By using the prior ethnic information, users can run JEPEGMIX for association summary data lacking allele frequency information. In this case, JEPEGMIX does not require study allele frequency information (the column ‘af1’) in the input file, but additionally JEPEGMIX needs one more input text file (population weight file) specifying ethnic proportion of the study samples. Here is a sample JEPEGMIX input file with 5 SNPs, which should be used when allele frequency information is not available.

rsid        chr  bp         a1  a2   z         info  

rs1000109   9    117908733  C   T   -0.714464  0.890

rs10001109  4    44904653   C   T    0.721919  0.731

rs1000112   22   28273339   C   G   -0.583666  0.445

rs10001127  4    130547376  T   C    0.069329  0.999

rs1000113   5    150240076  T   C   -1.447288  1.0

DISTMIX imputation: If your input summary data is generated from non-imputed genotype data (i.e. non-imputed GWASs or exome sequencing studies), you may want to impute summary statistics of unmeasured functional variants based on external panels (e.g. 1000 Genomes). By using "--impute" option, you can make the software impute them using DISTMIX before performing JEPEGMIX analysis. In this case, the software assumes that all SNPs in your input file are measured and does not use the imputation information provided in the "info" column.

Note: If the input summary data comes a different genome assembly (e.g. NCBI build 36 (hg18)) than NCBI build 37 (hg19), it needs to be converted by the user to NCBI build 37 (hg19) using a software, called liftover, from UCSC.

Here is a sample JEPEGMIX population weight file, which is needed when cohort allele frequency information is not available.

pop wgt   

ASW 0.1

GBR 0.3

FIN 0.2

CEU 0.25

IBS 0.15

The first line of the population weight file should be column names/headers. The file should contain two columns: 1) pop (population abbreviation) and 2) wgt (population proportion). The columns of data should be separated by white space. The weight file name should be specified with the ‘--populationWeight’ option.

JEPEGMIX output file format

The JEPEGMIX output file has nine columns: 1) gene name (geneid), 2) chi-square test statistic value (chisq), 3) degrees of freedom (df), 4) JEPEGMIX p-value (jepegmix_pval), 5) number of functional SNPs associated with gene (num_snp), 6) top functional category (top_categ), 7) top category p-value (top_categ_pval), 8) top functional SNP ID (top_snp) and 9) top SNP p-value (top_snp_pval). The first line of the output file is column names/headers. Here is a sample JEPEGMIX output file, with 5 genes:

geneid    chisq    df  jepegmix_pval   num_snp   top_categ   top_categ_pval   top_snp      top_snp_pval

FCGR3B    14.926   1   1.80E-05        2         PFS         1.10E-04         rs5030738    5.08E-05

CSPG4P2Y  14.774   2   3.29E-05        5         TRN         1.21E-04         rs10828664   5.55E-05

SLC7A7    19.902   3   3.40E-05        33        TRN         3.51E-04         rs9304810    1.07E-03

EDARADD   13.957   2   3.93E-05        2         PFS         1.86E-04         rs966365     7.80E-06

FAM71D    13.026   1   4.34E-05        26        TAR         3.07E-04         rs45515505   1.53E-04

Options

Getting started with examples

The users are required to download a tgz-compressed directory with pre-compiled JEPEGMIX executable (jepegmix), a sample input file (sample.input.txt) and a sample population weight file (sample.pop.wgt.txt) from the Download JEPEGMIX section above. The SNP annotation data set should be downloaded from Download SNP annotation data section. A tgz-compressed directory with 1000 Genomes reference panel datasets also should be downloaded from Download Reference Panels section. The tgz compressed files can be uncompressed using tar with the -zxvf options. Assume that the reference panel data file and its index file named as "1kg.geno.gz" and "1kg.index.gz" respectively are stored in a directory /path/to/reference/ and that the annotation dataset named as "jepeg.snp.annotations.v0.2.0.txt" and the sample population weight file (sample.pop.wgt.txt) are stored in directories /path/to/annotation/ and /path/to/weight/, respectively.

Scenario 1: Let's assume that the sample input data (association summary statistics) is generated from whole genome sequencing data. So, we want to perform JEPEGMIX analysis without DISTMIX imputation and store the results in a text file named as "jepegmix.output.txt". For this scenario, the following command can be used.

>> ./jepegmix sample.input.txt -o jepegmix.output.txt -i /path/to/reference/1kg.index.gz -r /path/to/reference/1kg.geno.gz -a /path/to/annotation/jepeg.snp.annotations.v0.2.0.txt -w /path/to/weight/sample.pop.wgt.txt

Scenario 2: If we want to impute unmeasured functional SNPs in our input data using DISTMIX before running JEPEGMIX analysis, we should use "--impute" option.

>> ./jepegmix --impute sample.input.txt -o jepegmix.output.txt -i /path/to/reference/1kg.index.gz -r /path/to/reference/1kg.geno.gz -a /path/to/annotation/jepeg.snp.annotations.v0.2.0.txt -w /path/to/weight/sample.pop.wgt.txt

Here, we didn't set the size of the prediction window and wing by using options "--windowSize" and "--wingSize", so JEPEGMIX will use the default size (1 Mb for the DISTMIX prediction window and 0.5 Mb for the wing).

Scenario 3: Here, we want to set the size of the DISTMIX prediction window (--windowSize or -n) to be 0.6 Mb and the size of the wing padded each side of the prediction window (--wingSize or -m) to be 0.2 Mb. For this scenario, we can use the following command.

>> ./jepegmix --impute sample.input.txt -o jepegmix.output.txt -i /path/to/reference/1kg.index.gz -r /path/to/reference/1kg.geno.gz  -a /path/to/annotation/jepeg.snp.annotations.v0.2.0.txt -w /path/to/weight/sample.pop.wgt.txt -n 0.6 -m 0.2