RAREMETAL Documentation

Useful Wiki Pages

There are a few pages in this Wiki that may be useful to rareMETAL users. Here are links to key pages:

• The RAREMETAL Home Page

• The RAREMETAL Quick Start Tutorial

• The RAREMETAL FAQ

• The RAREMETAL Command Reference

• The RAREMETALWORKER documentation

Key Features

RAREMETAL has the following features:

• RAREMETAL performs gene-based or region-based meta analysis using Burden tests with the following methods: CMC_counts, Madsen-Browning, SKAT, and Variable Threshold.
• RAREMETAL performs single variant metal-analysis by default.
• RAREMETAL allows customized groups of variants to be tested.
• RAREMETAL allows conditional analysis to be performed in both gene-level meta-analysis and single variants meta-analysis.
• RAREMETAL generate QQ plots and manhattan plots by default.

Brief Description

RAREMETAL is a computationally efficient tool for meta-analysis of rare variants using sequencing or genotyping array data. RAREMETAL takes summary statistics and LD matrices generated by RAREMETALWORKER, handles related and unrelated individuals, and supports both single variant and burden meta-analysis. RAREMETAL generated high quality plots by default and has options that allow users to build reports at different levels.

RAREMETAL is developed by Shuang Feng, Dajiang Liu and Gonçalo Abecasis. A R-package using the same methodology is available]. Manuscript for this tool is in preparation. Please contact sfengsph at umich dot edu for questions.

Approach

The key idea behind meta-analysis with RAREMETAL is that various gene-level test statistics can be reconstructed from single variant score statistics and that, when the linkage disequilibrium relationships between variants are known, the distribution of these gene-level statistics can be derived and used to evaluate signifi-cance. Single variant statistics are calculated using the Cochran-Mantel-Haenszel method. The main formulae are tabulated in the following:

Formulae for RAREMETAL

Test	Statistics	Null Distribution	Notation
Single Variant	$T=\sum _{i=1}^{n}{U_{i}}{\bigg /}{\sqrt {\sum _{i=1}^{n}{V_{i}}}}$	$T\sim \mathbf {N} (0,1)$	$U_{i}{\text{ is the score statistic from study }}i;$$V_{i}{\text{ is the variance of }}U_{i}.$
un-weighted Burden	$T_{b}=\sum _{i=1}^{n}{\mathbf {U_{i}} }{\Big /}{\sqrt {\sum _{i=1}^{n}{\mathbf {V_{i}} }}}$	$T_{b}\sim \mathbf {N} (0,1)$	$\mathbf {U_{i}} {\text{ is the vector of score statistics from study }}i,or$ $\mathbf {U_{i}} =\{U_{i1},...,U_{im}\};$ $\mathbf {V_{i}} {\text{ is the covariance of }}\mathbf {U_{i}} .$
Weighted Burden	$T_{wb}=\mathbf {w^{T}} \sum _{i=1}^{n}{\mathbf {U_{i}} }{\bigg /}{\sqrt {\mathbf {w^{T}} \left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right)\mathbf {w} }}$	$T_{wb}\sim \mathbf {N} (0,1)$	$\mathbf {w^{T}} =\{w_{1},w_{2},...,w_{m}\}^{T}{\text{ is the weight vector.}}$
VT	$T_{VT}=\max(T_{b\left(f_{1}\right)},T_{b\left(f_{2}\right)},\dots ,T_{b\left(f_{m}\right)}),{\text{ where}}$$T_{b\left(f_{j}\right)}={\boldsymbol {\phi }}_{f_{j}}^{\mathbf {T} }\sum _{i=1}^{n}{\mathbf {U_{i}} }{\bigg /}{\sqrt {{\boldsymbol {\phi }}_{f_{j}}^{\mathbf {T} }\left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right){\boldsymbol {\phi }}_{f_{j}}}}$	$\left(T_{b\left(f_{1}\right)},T_{b\left(f_{2}\right)},\dots ,T_{b\left(f_{m}\right)}\right)$$\sim \mathbf {MVN} \left(\mathbf {0} ,{\boldsymbol {\Omega }}\right){\text{,}}$${\text{where }}{\boldsymbol {\Omega _{ij}}}={\frac {{\boldsymbol {\phi }}_{f_{i}}^{T}\left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right){\boldsymbol {\phi }}_{f_{j}}}{{\sqrt {{\boldsymbol {\phi }}_{f_{i}}^{T}\left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right){\boldsymbol {\phi }}_{f_{i}}}}{\sqrt {{\boldsymbol {\phi }}_{f_{j}}^{T}\left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right){\boldsymbol {\phi }}_{f_{j}}}}}}$	${\boldsymbol {\phi }}_{f_{j}}{\text{ is a vector of }}0{\text{s and }}1{\text{s,}}$ ${\text{indicating the inclusion of a variant using threshold }}f_{j};$
SKAT	$\mathbf {Q} =\left(\sum _{i=1}^{n}{\mathbf {U_{i}^{T}} }\right)\mathbf {W} \left(\sum _{i=1}^{n}{\mathbf {U_{i}} }\right)$	$\mathbf {Q} \sim \sum _{i=1}^{m}{\lambda _{i}\chi _{1,i}^{2}},{\text{ where}}$ $\left(\lambda _{1},\lambda _{2},\dots ,\lambda _{m}\right){\text{ are eigen values of}}$$\left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right)^{\frac {1}{2}}\mathbf {W} \left(\sum _{i=1}^{n}{\mathbf {V_{i}} }\right)^{\frac {1}{2}}$	$\mathbf {W} {\text{ is a diagonal matrix of weights.}}$

Download and Installation

• University of Michigan CSG users can go to the following:

 /net/fantasia/home/sfengsph/code/Rare-Metal/raremetal/bin/raremetal 

Where to Download

• The software package for Linux and Mac (source code included) can be downloaded here: software package download

How to Compile

• Save it to your local path and decompress using the following command:

 tar xvzf raremetal.0.2.9.tar.gz

• Go to raremetal_0.2.9/raremetal/src and type the following command to compile:

 make

How to Execute

• Go to raremetal_0.2.9/raremetal/bin and use the following:

 ./raremetal

• For example usage, please refer to [example command lines]

Basic Usage Instructions

raeMETAL is a command line tool. It is typically run from a Linux or Unix prompt by invoking the command raremetal. In the following are descriptions of basic usage for meta analysis. A detailed TUTORIAL with toy data are also available.

Prepare Input Files

rareMETAL requires the following basic input files: summary statistics and covariance matrices of score statistics generated by rareMetalWorker, a file with list of studies to be included and a group file if gene-level meta-analysis is expected.

Summary Statistics

Files containing summary statistics and LD matrices generated by rareMetalWorker should be compressed and tabix indexed using the following commands:

bgzip study1.singlevar.score.txt
tabix -s 1 -b 2 -e 2 -c "#" study1.singlevar.score.txt.gz
bgzip study1.singlevar.cov.txt
tabix -s 1 -b 2 -e 2 -c "#" study1.singlevar.cov.txt.gz

List of Studies

• --studyName option is crucial for rareMETAL to work. Ignoring this option would lead to FATAL ERROR and rareMETAL would stop.
• The file should contain the path and prefix of the studies you want to include.
• If there is one or more studies that you want to excluded from your list, but want to save some effort of generating a new file, you can put a "#" in front of the line of record. rareMETAL would automatically exclude that study from meta analysis.
• An example file is in the following:

 /net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/TwinsUK/TwinsUK.TG
 #/net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/HUNT/RareMetalWorker/HUNT_MI_case.TG

• The above example study name file guides rareMETAL to look for summary statistics from TwinsUK study only, because "HUNT" study is commented out. The following two files are needed for rareMETAL to perform further analysis together with their tabix index file are needed.

 /net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/TwinsUK/TwinsUK.TG.singlevar.score.txt.gz	
 /net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/TwinsUK/TwinsUK.TG.singlevar.cov.txt.gz
 /net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/TwinsUK/TwinsUK.TG.singlevar.score.txt.gz.tbi
 /net/fantasia/home/sfengsph/prj/raremetal/raremetal/bin/META/TwinsUK/TwinsUK.TG.singlevar.cov.txt.gz.tbi

Group Rare Variants

From a Group File

• Grouping methods are only necessary when doing gene-based or group-based burden tests in meta-analysis.
• If none of the grouping method is specified, then only single variant meta-analysis will be performed.
• With --groupFile option, you can specify particular set of variants to be grouped for burden tests.
• The group file must be a tab or space delimited file in the following format:

 GROUP_ID MARKER1_ID MARKER2_ID MARKER3_ID ... 

• MARKER_ID must be in the following format:

 CHR:POS:REF:ALT

• An example group file is:

 PLEKHN1 1:901922:G:A    1:901923:C:A    1:902088:G:A    1:902128:C:T    1:902133:C:G    1:902176:C:T    1:905669:C:G        
 HES4    1:934735:A:C    1:934770:G:A    1:934801:C:T    1:935085:G:A    1:935089:C:G
 ISG15   1:949422:G:A    1:949491:G:A    1:949502:C:T    1:949608:G:A    1:949802:G:A    1:949832:G:A
 AGRN    1:970687:C:T    1:976963:A:G    1:977028:G:T    1:977356:C:T    1:977396:G:A    1:978628:C:T    1:978645:G:A             
 C1orf159        1:1021285:G:T   1:1021302:T:C   1:1021315:A:C   1:1021386:G:A   1:1022534:C:T   1:1025751:C:T   1:1026913:C:T

From an Annotated VCF File

If --groupFile option is NOT specified, rareMETAL will look for an annotated vcf file as blue print for variants to group. Users are also allowed to generate a vcf file based on the superset of variants from pooled samples, and annotate outside rareMETAL. Then, annotated vcf file can be used as input for rareMETAL for gene-level meta-analysis, or group files can be generated based on the annotated vcf file. Detailed description of these options are available. There are also examples of this usage at the bottom of this page.

QC Options

• rareMETAL allows filtering of variants from individual studies by their HWE pvalue and call rate, which are generated as part of the output from rareMetalWorker.
• To filter by HWE p-values, --hwe option should be used. The default is 0.0, which means not filtering any of the variants.
• To filter by call rate, --callRate option can be specified. The default is 0.0, which allows no filtering utilized.

Association Options

• Currently, CMC type burden test, Madsen-Browning burden test, Variable Threshold burden test and SKAT are provided in rareMETAL, by specifying --burden, --MB, --VT and --SKAT.
• --maf specifies the minor allele frequency cutoff when doing gene-based or group-based burden tests. The default is maf<0.05.

Conditional Analysis

• To decide whether a signal is caused by shadowing a significant common variant nearby, rareMETAL also enables conditional analysis with a list of variants to be conditioned upon provided in a file as input for --condition option. An example input file should be space or tab delimited as in the following. When alleles do not match the ref and alt alleles from samples, the variant will be skipped from conditional analysis.

1:861349:C:T 1:905901:G:A 20:986998:G:C 22:3670691:A:G

Additional Analysis Options

Group Rare Variants from Annotated VCF

• If --groupFile option is NOT specified, RAREMETAL will look for an annotated vcf file as blue print for variants to group.
• The annotated VCF file should be specified using --annotatedVcf option.
• --annotation should be used with --annotatedVcf together when specific category of functional variants are of interest to be grouped. For example, if grouping nonsynonymous and splicing variants are of interests, the following should be included in command line:

 --annotatedVcf your.annotated.vcf --annotation nonsyn/splicing
 Note: this allows you to group variants that are annotated starting with nonsyn or splicing (not case-sensitive).

• Special format for the annotated VCF file is required: all annotation information should be coded in INFO field in VCF file, starting with the key "ANNO=". An example annotated VCF file is in the following:

 #CHROM    POS     ID      REF     ALT     QUAL    FILTER  INFO
 1       19208194        .       G       A       100     PASS      
 AC=3;ANNO=nonsynonymous:ALDH4A1:NM_170726:exon8:c.C866T:p.P289L,ALDH4A1:NM_001161504:exon8:c.C686T:p.P229L,ALDH4A1:NM_003748:exon8:c.C866T:p.P289L,;
 ANNO=splicing:ALDH4A1
 1       19208293        .       G       C       100     PASS    AC=7;STUDIES=5;MAC=7;MAF=0.001;DESIGN=TBD_ASSAY;DSCORE=1.00;
 ANNO=nonsynonymous:ALDH4A1:NM_170726:exon8:c.C767G:p.P256R,ALDH4A1:NM_001161504:exon8:c.C587G:p.P196R,ALDH4A1:NM_003748:exon8:c.C767G:p.P256R,

• Notice that each variant is allowed to have more than one annotations; but each annotation should start with a new key "ANNO=" followed by annotation:genename:other transcript information.

Generate a VCF File to Annotate Outside of Rare Metal

• --writeVCF allows user to write a VCF file including pooled single variants from all studies. Then users can use their favorite annotation tool to annotate the VCF file. After annotating the VCF file, users can use that file as input for RAREMETAL for further gene-based or region-based meta analysis.
• The output vcf file will be name as: yourPrefix.pooled.variants.vcf. An example output vcf file is in the following:

 #CHROM    POS     ID      REF     ALT     QUAL    FILTER  INFO
 1       115658497       115658497       G       A       .       .       ALT_AF=0.380906;
 2       74688884        74688884        G       A       .       .       ALT_AF=8.33611e-05;
 3       121414217       121414217       C       A       .       .       ALT_AF=0.0747833;

Options for Report Generation

• --correctGC generates QQ plots and manhattan plots with pvalues corrected using genomic control.
• --prefix allows customized prefix for output files.
• --longOutput allows users to output not only burden test results but also the single variant results (allele frequencies, effect sizes, and p-values) for the variants being grouped together. Please refer to the output files section for detailed explanation and examples.
• --tabulateHits works with --hitsCutoff together to generate reports for genes that have p-value less than specified cutoff from burden tests or SKAT. The default cutoff of p-value for genes to be reported is 1.0e-06, which can be specified by --hitsCutoff option. For more explanations and examples, please go to Tabulated Hits.

Miscellaneous Options

• --tabix allows rapid analysis when number of groups/genes of interests are small. Currently, when number of groups is less than 100, --tabix option is automatically turned on.

Reports Generated by RAREMETAL

Single Variant Meta Analysis Output

TABLES

• Single variant meta analysis output has the following components: header, results and footnote.
• Header lines start with "##" shows summary of the meta analysis including method used, number of studies, and total sample size.
• Header line starts with "#" are column headers for results table.
• Footnote also starts with "#", where genomic controls from each study and the overall sample are reported.
• An example single variant meta analysis output is shown below:

 ##Method=SinglevarScore
 ##STUDY_NUM=2
 ##TotalSampleSize=14308
 #CHROM  POS     REF     ALT     POOLED_ALT_AF   EFFECT_SIZE     DIRECTION_BY_STUDY      PVALUE
 1       115658497       G       A       0.380906        0.00954332      ++      0.45828
 2       74688884        G       A       8.33611e-05     -0.196387       -!      0.845372
 3       121414217       C       A       0.0747833       0.0216982       -+      0.34453
 6       137245814       G       C       0.000803746     0.105693        ++      0.601805

• A detailed explanation of each column is in the following:

 CHROM:              Chromosome Name
 POS:                Variant Position
 REF:                Reference Allele Label
 ALT:                Alternative Allele Label
 POOLED_ALT_AF:      Pooled Alternative Allele Frequency
 EFFECT_SIZE:        Alternative Allele Effect Size
 DIRECTION_BY_STUDY: Effect size direction of alternative allele from each study. 
                     The order of study is consistent with the order of studies listed in the input file for option --studyName. 
                     "?" means the variant is not observed or monomorphic from the study. 
                     "!" means the variant observed from this study has different alleles from those in the first study.

PLOTS

RAREMETAL generates QQ plots and manhattan plots from single variant meta-analysis by default. Three QQ plots are generated, one with all variants included, one of variants with maf<0.05 and one of variants with maf<0.01. All plots are saved in a pdf file named yourPrefix.meta.plots.pdf. Genomic controls are also reported in the title of plots. When --correctGC option is specified, GC corrected plots are also generated.

Gene-level Tests Meta-Analysis Output

LONG TABLES

When --longOutput is used, output includes both burden test results of genes and single variant results of the variants included in burden tests. Here is an example of output file from SKAT when --longOutput is specified.

 ##Method=Burden
 ##STUDY_NUM=2
 ##TotalSampleSize=14308
 #GROUPNAME      NUM_VAR VARs    MAFs    SINGLEVAR_EFFECTs       SINGLEVAR_PVALUEs       AVG_AF  MIN_AF  MAX_AF  EFFECT_SIZE     PVALUE
 NOC2L   7       1:880502:C:T;1:881918:G:A;1:887799:C:T;1:888659:T:C;1:889238:G:A;1:891591:C:T;1:892380:G:A        0.000166722,0.0242172,0.0109203,0.0355845,0.0333729,0.00700233,0.00200067       -0.183575,-0.00228307,-0.0598337,0.0220595,0.0229464,-0.0302768,-0.0200417      0.790161,0.953446,0.515806,0.503548,0.499251,0.791773,0.926625  0.0161807       0.000166722     0.0355845       0.00667875      0.662531
 KLHL17  2       1:897285:A:G;1:898869:C:T       0.0148408,0.00108369    -0.0502034,-0.0256403   0.528269,0.934606       0.00796222      0.00108369      0.0148408       -0.0484494      0.528878

SHORT TABLES

Otherwise, single variant results of variants included in burden tests will not be included in the output. Here is an example of output file from SKAT when --longOutput is not specified.

 ##Method=Burden
 ##STUDY_NUM=2
 ##TotalSampleSize=14308
 #GROUPNAME      NUM_VAR VARs    AVG_AF  MIN_AF  MAX_AF  EFFECT_SIZE     PVALUE
 NOC2L   7       1:880502:C:T;1:881918:G:A;1:887799:C:T;1:888659:T:C;1:889238:G:A;1:891591:C:T;1:892380:G:A      0.0161807       0.000166722     0.0355845       0.00667875      0.662531
 KLHL17  2       1:897285:A:G;1:898869:C:T       0.00796222      0.00108369      0.0148408       -0.0484494      0.528878

TABULATED HITS

• When --tabulateHits is specified, top hits from Burden tests will be generated. Each method will have an individual tabulated file generated. The purpose of this tabulated file is to list burden test results of top hits together with single variant results from variants being grouped in burden tests. The difference between this file and the standard long-format output file from burden test is that each row of the file represents a single variant that is included in the gene for burden test. This format allows each sorting on users end.

• Tabulated top hits are saved in the file:

 yourPrefix.meta.tophits.youMethod.tbl (example files names: TG.meta.tophits.burden.tbl, LDL.meta.tophits.SKAT.tbl)

• The following items are tabulated in the output:

 GENE: Gene name.
 METHOD: Burden test used.
 GENE_PVALUE: P-value from gene-based burden tests.
 MAF_CUTOFF: MAF cutoff used when doing gene-based tests.
 ACTUAL_CUTOFF: Actual MAF cutoff used. (This will be different from MAF_CUTOFF only for Variable Threshold method.
                Otherwise, it will be the same as MAF_CUTOFF.)
 VAR: Variant name in CHR:POS:REF:ALT format.
 MAF: Single variant pooled MAF from all samples.
 EFFSIZE: Effect size from single variant meta analysis. 
 PVALUE: Pvalue from single variant meta analysis.

• An example of tabulated hits from a standard burden test with maf<0.05 as criterion is shown in the following:

 GENE    METHOD  GENE_PVALUE     MAF_CUTOFF      ACTUAL_CUTOFF   VARS    MAFS    EFFSIZES        PVALUES
 PCSK9   BURDEN_0.050    7.54587e-11     0.05    0.05    1:55505647:G:T  0.0396631       -0.442192       2.10159e-46
 PCSK9   BURDEN_0.050    7.54587e-11     0.05    0.05    1:55518371:G:A  0.0237138       0.0548733       0.430246
 PCSK9   BURDEN_0.050    7.54587e-11     0.05    0.05    1:55529187:G:A  0.0433324       0.0946321       0.00129942
 APOE    BURDEN_0.050    2.83457e-72     0.05    0.05    19:45412079:C:T 0.0413056       -0.554561       2.83457e-72

• According to the example above, PCSK9 had a p-value of 7.54587e-11 from the gene-based burden test, where three variants from this gene were included. Another hit from this meta analysis is APOE, where only one variant was included in the burden test.

PLOTS

RAREMETAL generates QQ plots and manhattan plots from single variant and gene-level meta-analysis by default. Example QQ plots and manhattan plots are:

LOG

• A log file is automatically generated by RAREMETAL to save the parameters in effect. An example is in the following:

 The following parameters are in effect:
 
 List of Studies:
 ============================
 --studyName [studyName.SardiNia]
 
 Grouping Methods:
 ============================
 --groupFile [genes.file]
 --annotatedVcf []
 --annotation []
 --writeVcf [OFF]
 
 QC Options:
 ============================
 --hwe [0]
 --callRate [0] 
 
 Association Methods:
 ============================
 --burden [true]
 --MB [false]
 --SKAT [false]
 --VT [false]
 --condition [condition.file]
 
 Other Options:
 ============================
 --tabix [OFF]
 --correctGC [ON]
 --prefix [test]
 --maf [0.05]
 --longOutput [false]
 --tabulateHits [false]
 --hitsCutoff [1e-06]

Example Command lines

• Here is an example command line to do single variant meta analysis only:

 ./raremetal --studyName your.studyName.file --prefix yourPrefix 

• When you want to do all burden tests using a group file to specify which variants to group:

 ./raremetal --studyName your.studyName.file --groupFile your.groupfile --burden --MB --SKAT --VT --maf 0.01 --prefix yourPrefix
 (NOTE: this will generate single variant meta analysis result and the short format output for burden test results.)

• Here is how to do all SKAT meta analysis using a group file and request a long format output together with tabulated hits:

 ./raremetal --studyName your.studyName.file --groupFile your.groupfile --SKAT --longOutput --tabulateHits --hitsCutoff 1.0e-07 --prefix yourPrefix

• Here is an example of adding QC filters to variants when doing meta analysis.

 ./raremetal --studyName your.studyName.file --groupFile your.groupfile --SKAT --longOutput --tabulateHits --hitsCutoff 1.0e-07 --hwe 1e-06 --callRate 0.98 --prefix yourPrefix

• Here is how to do the same thing but reading grouping information from an annotated VCF file:

 ./raremetal --studyName your.studyName.file --annotatedVcf your.annotated.vcf --annotation nonsyn/stop/splicing --SKAT --longOutput --tabulateHits --hitsCutoff 1.0e-07 --hwe 1e-06 --callRate 0.98 --prefix yourPrefix

• If you want to write a VCF file of pooled variants from all studies, annotate them using your favorite annotation program, and then come back to RAREMETAL with the annotate VCF file to do burden tests:

 First, use the following command to write the VCF file:
 ./raremetal --studyName your.studyName.file --writeVcf --prefix yourPrefix
 Second, annotate the VCF file using your favorite annotation program. (Annotated VCF file has to follow the format described here: annotated VCF format)
 Third, use the following command to do meta analysis:
 ./raremetal --studyName your.studyName.file --annotatedVcf your.annotated.vcf --annotation nonsyn/splicing/stop --burden --MB --SKAT --VT --maf 0.01 --prefix yourPrefix

TUTORIAL

• For a comprehensive tutorial of RAREMETALWORKER and RAREMETAL using example data sets, please go to the following:

 RAREMETAL Tutorial

CONTACT

Please email Shuang Feng (sfengsph at umich dot edu) for questions.

Change Log

• Version 0.0.1 released to U of M CSG group. (2/13/2013)
• Version 0.0.1 released to public. (2/24/2013)
• Version 0.1.2 released to public after fixing a few bugs, adding conditional analysis and automatic graphing to the tool. (8/5/2013)