From shenxia911 at gmail.com  Sat Apr 16 22:33:00 2016
From: shenxia911 at gmail.com (Xia Shen)
Date: Sat, 16 Apr 2016 22:33:00 +0200
Subject: [GenABEL-dev] patch for MultiABEL load.summary
In-Reply-To: <2940BD3C-48AF-4804-83AE-8DE892C14592@gmail.com>
References: <2940BD3C-48AF-4804-83AE-8DE892C14592@gmail.com>
Message-ID: <16B91094-33EF-46A9-B6B0-610BFEE10288@gmail.com>

Thanks for the patch! Please commit to the SVN, and thereafter we can do more testing.

Xia

> On 23 Mar 2016, at 13:24, Yurii Aulchenko <yurii.aulchenko at gmail.com> wrote:
> 
> Dear Xia,
> 
> I made a patch for MultiABEL::load.summary function, which extends functionality in two ways:
> 
> 1) allows the user to specify what input file column names correspond to required columns
> 
> 2) adds option fixedN, which, when provided is used as the sample size
> 
> The patch is attached. Although this all seems to work fine on my test examples, my coding skills degraded in last years, so it will be worth checking the code. 
> 
> Let me know then if you?d like to apply the patch yourself; alternatively I can commit my changes to SVN directly.
> 
> best,
> Yurii
> 
> <patch.txt>


From yurii.aulchenko at gmail.com  Sat Apr 16 22:41:13 2016
From: yurii.aulchenko at gmail.com (Yurii Aulchenko)
Date: Sat, 16 Apr 2016 22:41:13 +0200
Subject: [GenABEL-dev] patch for MultiABEL load.summary
In-Reply-To: <16B91094-33EF-46A9-B6B0-610BFEE10288@gmail.com>
References: <2940BD3C-48AF-4804-83AE-8DE892C14592@gmail.com>
 <16B91094-33EF-46A9-B6B0-610BFEE10288@gmail.com>
Message-ID: <99E270C8-4981-42A5-876F-5B150550C930@gmail.com>

great! - committed as r2052

Yurii

> On 16 Apr 2016, at 22:33, Xia Shen <shenxia911 at gmail.com> wrote:
> 
> Thanks for the patch! Please commit to the SVN, and thereafter we can do more testing.
> 
> Xia
> 
>> On 23 Mar 2016, at 13:24, Yurii Aulchenko <yurii.aulchenko at gmail.com> wrote:
>> 
>> Dear Xia,
>> 
>> I made a patch for MultiABEL::load.summary function, which extends functionality in two ways:
>> 
>> 1) allows the user to specify what input file column names correspond to required columns
>> 
>> 2) adds option fixedN, which, when provided is used as the sample size
>> 
>> The patch is attached. Although this all seems to work fine on my test examples, my coding skills degraded in last years, so it will be worth checking the code. 
>> 
>> Let me know then if you?d like to apply the patch yourself; alternatively I can commit my changes to SVN directly.
>> 
>> best,
>> Yurii
>> 
>> <patch.txt>
> 


From yurii.aulchenko at gmail.com  Sat Apr 16 22:44:28 2016
From: yurii.aulchenko at gmail.com (Yurii Aulchenko)
Date: Sat, 16 Apr 2016 22:44:28 +0200
Subject: [GenABEL-dev] [Genabel-commits] r2052 - in pkg/MultiABEL: . R
	man src
In-Reply-To: <20160416203943.9E3CF1864C9@r-forge.r-project.org>
References: <20160416203943.9E3CF1864C9@r-forge.r-project.org>
Message-ID: <C205EFA3-229E-4611-8D4C-1224A1AAC6FA@gmail.com>

Xia, I only changed load.summary.R and then re-build documentation (consecutively load.summary.Rd was changed)

All other ?changes' are probably due to differences how roxygen2 worked on my system; did not touch other functions; should have avoided to commit these, but in principle this should not matter

best,
Yurii

> On 16 Apr 2016, at 22:39, noreply at r-forge.r-project.org wrote:
> 
> Author: yurii
> Date: 2016-04-16 22:39:42 +0200 (Sat, 16 Apr 2016)
> New Revision: 2052
> 
> Modified:
>   pkg/MultiABEL/DESCRIPTION
>   pkg/MultiABEL/R/load.summary.R
>   pkg/MultiABEL/man/MultiABEL.Rd
>   pkg/MultiABEL/man/MultiLoad.Rd
>   pkg/MultiABEL/man/MultiMeta.Rd
>   pkg/MultiABEL/man/MultiRep.Rd
>   pkg/MultiABEL/man/MultiSummary.Rd
>   pkg/MultiABEL/man/Multivariate.Rd
>   pkg/MultiABEL/man/load.summary.Rd
>   pkg/MultiABEL/src/symbols.rds
> Log:
> added options columnNames and fixedN to load.summary
> 
> Modified: pkg/MultiABEL/DESCRIPTION
> ===================================================================
> --- pkg/MultiABEL/DESCRIPTION	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/DESCRIPTION	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -5,7 +5,9 @@
> Date: 2016-02-25
> Author: Xia Shen
> Maintainer: Xia Shen <xia.shen at ki.se>
> -Description: Multivariate genome-wide association analyses. The analysis can be performed on individual-level data or multiple single-trait genome-wide summary statistics. 
> +Description: Multivariate genome-wide association analyses. The analysis can be
> +    performed on individual-level data or multiple single-trait genome-wide summary
> +    statistics.
> Depends:
>     R (>= 2.10),
>     svMisc
> @@ -15,3 +17,4 @@
> License: GPL (>= 2)
> LazyLoad: yes
> Packaged: 2016-02-25 15:52:58 CET; xia
> +RoxygenNote: 5.0.1
> 
> Modified: pkg/MultiABEL/R/load.summary.R
> ===================================================================
> --- pkg/MultiABEL/R/load.summary.R	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/R/load.summary.R	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -27,11 +27,15 @@
> #' are available, for which the argument \code{vars} has to be given. 
> #' @param vars A numeric vector gives the variance of the genotypes at each SNP, e.g. coded as 0, 1 and 2.
> #' Only used when \code{type = "precise"}.
> +#' @param columnNames A vector with names of columns containing necessary information in the input file;
> +#' default values are c('snp','a1','freq','beta','se','n'). The values are case-insensitive. 
> +#' @param fixedN sample size to assume across all analyses, when provided, this number will be used
> +#' (instead of the ones specified in the input files)
> #' 
> #' @return The function returns a list of class \code{multi.summary}, containing two elements: \code{gwa}
> #' (the cleaned data to be processed in multi-trait GWAS) and \code{cor.pheno} (user input or estimated).
> #' 
> -#' @author Xia Shen
> +#' @author Xia Shen, Yurii Aulchenko
> #' 
> #' @references 
> #' Xia Shen, Zheng Ning, Yakov Tsepilov, Masoud Shirali, 
> @@ -67,7 +71,8 @@
> #' @aliases load.summary
> #' @keywords multivariate, meta-analysis
> #' 
> -`load.summary` <- function(files, cor.pheno = NULL, indep.snps = NULL, est.var = FALSE, type = 'outbred', vars = NULL) {
> +`load.summary` <- function(files, cor.pheno = NULL, indep.snps = NULL, est.var = FALSE, type = 'outbred', vars = NULL,
> +                           columnNames = c ('snp','a1','freq','beta','se','n'), fixedN = NULL ) {
> 	if (!all(is.character(files))) {
> 		stop('files should be given as strings!')
> 	}
> @@ -83,6 +88,30 @@
> 			stop('wrong dimensions of cor.pheno!')
> 		}
> 	}
> +        columnNames <- tolower( columnNames)
> +        if (!is.null( fixedN )) if (fixedN <= 0) {
> +            stop('fixedN should be a positive number')
> +        }
> +        if (is.null(fixedN)) { colNamLen = 6 } else { colNamLen = 5 }
> +        if (!is.character(columnNames)) {
> +            stop('columnNames should be character')
> +        }
> +        if (length( columnNames ) != colNamLen) {
> +            cat('columnNames should be a vector with',colNamLen,'elements')
> +            stop('... exiting')
> +        }
> +        if ( length(unique(columnNames)) != colNamLen ) {
> +            stop('elements of columnNames must be unique')
> +        }
> +        # column Names Translation
> +        cNT = list(
> +            'snp' = columnNames[1],
> +            'a1'  = columnNames[2],
> +            'freq'= columnNames[3],
> +            'beta'= columnNames[4],
> +            'se'  = columnNames[5],
> +            'n'   = columnNames[6]
> +            )
> 	cat('loading data ...\n')
> 	data <- c()
> 	fn <- files # rev(files)
> @@ -90,17 +119,27 @@
> 	for (i in m:1) {
> 		dd <- read.table(fn[i], header = TRUE, stringsAsFactors = FALSE)
> 		colnames(dd) <- tolower(colnames(dd))
> -		idx <- which(duplicated(dd$snp))
> +                currentColNames <- colnames(dd)
> +                if ( any( !( columnNames %in% currentColNames ) ) ) {
> +                    cat('file column names do not match columnNames in ',fn[i],'... ')
> +                    stop('exiting')
> +                }
> +		idx <- which(duplicated(dd[, cNT[['snp']] ]))
> 		if (length(idx) > 0) {
> 			data[[i]] <- dd[-idx,]
> -			rownames(data[[i]]) <- dd$snp[-idx]
> +			rownames(data[[i]]) <- dd[ -idx , cNT[['snp']] ]
> 		} else {
> 			data[[i]] <- dd
> -			rownames(data[[i]]) <- dd$snp
> +			rownames(data[[i]]) <- dd[, cNT[['snp']] ]
> 		}
> 		if (est.var) {
> -			D <-  dd$n*2*dd$freq*(1 - dd$freq)
> -			vy <- D*dd$se**2 + D*dd$beta**2/(dd$n - 1)
> +                        if (!is.null(fixedN)) {
> +                            D <-  dd[, cNT[['n']] ]*2*dd[, cNT[['freq']] ]*(1 - dd[, cNT[['freq']] ])
> +                            vy <- D*dd[ , cNT[['se']] ]**2 + D*dd[, cNT[['beta']] ]**2/(dd[, cNT[['n']] ] - 1)
> +                        } else {
> +                            D <-  fixedN*2*dd[, cNT[['freq']] ]*(1 - dd[, cNT[['freq']] ])
> +                            vy <- D*dd[ , cNT[['se']] ]**2 + D*dd[, cNT[['beta']] ]**2/( fixedN - 1)
> +                        }
> 			dvy <- density(na.omit(vy))
> 			vys[i] <- dvy$x[which.max(dvy$y)] #median(vy, na.rm = TRUE)
> 		}
> @@ -109,9 +148,9 @@
> 	cat('\n')
> 	if (est.var) cat('phenotypic variances are:', vys, '\n')
> 	cat('checking markers ...\n')
> -	snps <- data[[1]]$snp
> +	snps <- data[[1]][, cNT[['snp']] ]
> 	for (i in 2:m) {
> -		snps <- data[[i]]$snp[data[[i]]$snp %in% snps]
> +		snps <- data[[i]][ data[[i]][, cNT[['snp']] ] %in% snps, cNT[['snp']] ]
> 		progress(i/m*100)
> 	}
> 	snps <- unique(snps)
> @@ -124,31 +163,37 @@
> 	cat('\n')
> 	cat('correcting parameters ...\n')
> 	for (i in 2:m) {
> -		if (any(data[[i]]$a1 != data[[1]]$a1)) {
> -			adj <- 2*as.numeric(data[[i]]$a1 == data[[1]]$a1) - 1
> -			data[[i]]$beta <- data[[i]]$beta*adj
> -			data[[i]]$freq <- (adj == 1)*data[[i]]$freq + (adj == -1)*(1 - data[[i]]$freq)
> +		if (any( data[[i]][, cNT[['a1']] ] != data[[1]][, cNT[['a1']] ] )) {
> +			adj <- 2*as.numeric( data[[i]][, cNT[['a1']] ] == data[[1]][, cNT[['a1']] ] ) - 1
> +			data[[i]][, cNT$beta ] <- data[[i]][, cNT$beta ]*adj
> +			data[[i]][, cNT$freq ] <- (adj == 1)*data[[i]][, cNT$freq] + (adj == -1)*(1 - data[[i]][,cNT$freq])
> 		}
> 		progress(i/m*100)
> 	}
> 	cat('\n')
> 	cat('adjusting sample size ... ')
> 	n0 <- matrix(NA, nrow(data[[1]]), m)
> -	for (i in 1:m) {
> -		n0[,i] <- data[[i]]$n
> -	}
> +        if (is.null(fixedN)) {
> +            for (i in 1:m) {
> +		n0[,i] <- data[[i]][,cNT$n]
> +            }
> +        } else {
> +            for (i in 1:m) {
> +		n0[,i] <- fixedN
> +            }            
> +        }
> 	n <- apply(n0, 1, "min")
> 	cat('done.\n')
> 	cat('finalizing summary statistics ...\n')
> 	gwa0 <- matrix(NA, nrow(data[[1]]), 2*m + 2)
> 	for (i in 1:m) {
> -		gwa0[,i*2 - 1] <- data[[i]][,'beta']
> -		gwa0[,i*2] <- data[[i]][,'se']
> +		gwa0[,i*2 - 1] <- data[[i]][,cNT$beta]
> +		gwa0[,i*2] <- data[[i]][,cNT$se]
> 		progress(i/m*100)
> 	}
> -	gwa0[,2*length(data) + 1] <- data[[1]][,'freq']
> +	gwa0[,2*length(data) + 1] <- data[[1]][,cNT$freq]
> 	gwa0[,2*length(data) + 2] <- n
> -	rownames(gwa0) <- data[[1]]$snp
> +	rownames(gwa0) <- data[[1]][,cNT$snp]
> 	gwa0 <- na.omit(gwa0)
> 	cat('\n')
> 	if (is.null(cor.pheno)) {
> 
> Modified: pkg/MultiABEL/man/MultiABEL.Rd
> ===================================================================
> --- pkg/MultiABEL/man/MultiABEL.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/MultiABEL.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/MultiABEL.R
> \docType{package}
> \name{MultiABEL}
> @@ -10,16 +10,16 @@
> MultiABEL: Multivariate Genome-Wide Association Analyses
> }
> \details{
> -Performing multivariate genome-wide association (MVGWA) analyses.
> +Performing multivariate genome-wide association (MVGWA) analyses. 
> The modules are compatible with existing *ABEL data formats. The GWA
> analyses can be done on individual level data or on
> single-trait GWA summary statistics only.
> 
> For converting data from other formats, see
> 
> -\code{\link{convert.snp.illumina}} (Illumina/Affymetrix-like format). This is
> -our preferred converting function, very extensively tested. Other conversion
> -functions include:
> +\code{\link{convert.snp.illumina}} (Illumina/Affymetrix-like format). This is 
> +our preferred converting function, very extensively tested. Other conversion 
> +functions include: 
> \code{\link{convert.snp.text}} (conversion from human-readable GenABEL format),
> \code{\link{convert.snp.ped}} (Linkage, Merlin, Mach, and similar files),
> \code{\link{convert.snp.mach}} (Mach-format),
> @@ -27,16 +27,16 @@
> \code{\link{convert.snp.affymetrix}} (BRML-style files).
> 
> For converting of GenABEL's data to other formats, see
> -\code{\link{export.merlin}} (MERLIN and MACH formats),
> +\code{\link{export.merlin}} (MERLIN and MACH formats), 
> \code{\link{export.impute}} (IMPUTE, SNPTEST and CHIAMO formats),
> \code{\link{export.plink}} (PLINK format, also exports phenotypic data).
> 
> To load the data, see \code{\link{load.gwaa.data}}.
> 
> -For conversion to DatABEL format (used by ProbABEL and some other
> -GenABEL suite packages), see
> -\code{\link{impute2databel}},
> -\code{\link{impute2mach}},
> +For conversion to DatABEL format (used by ProbABEL and some other 
> +GenABEL suite packages), see 
> +\code{\link{impute2databel}}, 
> +\code{\link{impute2mach}}, 
> \code{\link{mach2databel}}.
> 
> For data managment and manipulations see
> 
> Modified: pkg/MultiABEL/man/MultiLoad.Rd
> ===================================================================
> --- pkg/MultiABEL/man/MultiLoad.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/MultiLoad.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/MultiLoad.R
> \name{MultiLoad}
> \alias{MultiLoad}
> @@ -7,7 +7,8 @@
> \title{Load individual-level data for multivariate GWA analysis}
> \usage{
> MultiLoad(gwaa.data = NULL, phenofile = NULL, genofile = NULL, trait.cols,
> -  covariate.cols = NULL, cuts = 20, impute = TRUE, ...)
> +  covariate.cols = NULL, cuts = 20, impute = TRUE, gaussianize = TRUE,
> +  ...)
> }
> \arguments{
> \item{gwaa.data}{An (optional) object of \code{\link{gwaa.data-class}}.}
> @@ -36,7 +37,7 @@
> to perform multivariate test for each genetic variant.
> }
> \note{
> -Either \code{gwaa.data} (for GenABEL data format) or the combination of
> +Either \code{gwaa.data} (for GenABEL data format) or the combination of 
> \code{phenofile} and \code{genofile} (for DatABEL data format) has to be provided.
> If all are provided, only \code{phenofile} and \code{genofile} will be used. When using
> DatABEL format input, individual IDs in \code{phenofile} and \code{genofile} have to match!
> @@ -48,18 +49,18 @@
> data(ge03d2ex.clean)
> 
> ## running multivariate GWAS for 3 traits: height, weight, bmi
> -loaded <- MultiLoad(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8),
> +loaded <- MultiLoad(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8), 
>                     covariate.cols = c(2, 3))
> 
> ## converting the same dataset into DatABEL format files
> require(DatABEL)
> -write.table(phdata(ge03d2ex.clean), 'pheno.txt', col.names = TRUE, row.names = TRUE,
> +write.table(phdata(ge03d2ex.clean), 'pheno.txt', col.names = TRUE, row.names = TRUE, 
>             quote = FALSE, sep = '\\t')
> geno <- as.double(ge03d2ex.clean)
> matrix2databel(geno, 'geno')
> 
> ## running the multivariate GWAS again
> -loaded <- MultiLoad(phenofile = 'pheno.txt', genofile = 'geno', trait.cols = c(5, 6, 8),
> +loaded <- MultiLoad(phenofile = 'pheno.txt', genofile = 'geno', trait.cols = c(5, 6, 8), 
>                     covariate.cols = c(2, 3))
> }
> }
> @@ -68,7 +69,7 @@
> }
> \references{
> Xia Shen, ..., Jim Wilson, Gordan Lauc, Yurii Aulchenko (2015).
> -Multi-omic-variate analysis identified novel loci associated with
> +Multi-omic-variate analysis identified novel loci associated with 
> compound N-Glycosylation of human Immunoglobulin G. \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/man/MultiMeta.Rd
> ===================================================================
> --- pkg/MultiABEL/man/MultiMeta.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/MultiMeta.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/MultiMeta.R
> \name{MultiMeta}
> \alias{MultiMeta}
> @@ -17,7 +17,7 @@
> \value{
> The function returns a matrix containing the meta-analysis results, where the row names are
> the variants names, and the column names are the names of the studies provided in \code{reslist} or
> -generated by the program if no names are given, with an extra column \code{"p.meta"} containing the
> +generated by the program if no names are given, with an extra column \code{"p.meta"} containing the 
> meta-analysis P-values. The results are also written into \code{outfile}.
> }
> \description{
> @@ -30,9 +30,9 @@
> data(ge03d2ex)
> 
> ## in each dataset, running multivariate GWAS for 3 traits: height, weight, bmi
> -res1 <- Multivariate(gwaa.data = ge03d2, trait.cols = c(5, 6, 8),
> +res1 <- Multivariate(gwaa.data = ge03d2, trait.cols = c(5, 6, 8), 
>                      covariate.cols = c(2, 3))
> -res2 <- Multivariate(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8),
> +res2 <- Multivariate(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8), 
>                      covariate.cols = c(2, 3))
> 
> ## running meta-analysis by combining the P-values
> @@ -44,7 +44,7 @@
> }
> \references{
> Xia Shen, ..., Gordan Lauc, Jim Wilson, Yurii Aulchenko (2014).
> -Multi-omic-variate analysis identified the association between 14q32.33 and
> +Multi-omic-variate analysis identified the association between 14q32.33 and 
> compound N-Glycosylation of human Immunoglobulin G \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/man/MultiRep.Rd
> ===================================================================
> --- pkg/MultiABEL/man/MultiRep.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/MultiRep.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/MultiRep.R
> \name{MultiRep}
> \alias{MultiRep}
> @@ -14,30 +14,30 @@
> \item{training.pheno}{An (optional) matrix or data frame contains the phenotype data for the discovery
> sample, preferrably adjusted for fixed effects and population structure before multivariate GWA analysis.}
> 
> -\item{training.phenofile}{An (optional) plain text file contains phenotypes for the discovery sample.
> +\item{training.phenofile}{An (optional) plain text file contains phenotypes for the discovery sample. 
> If this is provided, it will serve as \code{training.pheno}.}
> 
> \item{test.pheno}{An (optional) matrix or data frame contains the phenotype data for the replication
> sample, preferrably adjusted for fixed effects and population structure.}
> 
> -\item{test.phenofile}{An (optional) plain text file contains phenotypes of the replication sample.
> +\item{test.phenofile}{An (optional) plain text file contains phenotypes of the replication sample. 
> If this is provided, it will serve as \code{test.pheno}.}
> 
> \item{pheno.names}{A vector (length > 1) giving the column names of the phenotypes to be analyzed.}
> 
> -\item{training.geno}{A matrix or data.frame that contains the discovery sample genotype dosages
> +\item{training.geno}{A matrix or data.frame that contains the discovery sample genotype dosages 
> of the variants to replicate.}
> 
> -\item{test.geno}{A matrix or data.frame that contains the replication sample genotype dosages
> -of the variants to replicate. This object should have the same column names and order
> +\item{test.geno}{A matrix or data.frame that contains the replication sample genotype dosages 
> +of the variants to replicate. This object should have the same column names and order 
> as \code{training.geno}.}
> }
> \value{
> -The function returns a list of 3 matrices. \code{$replication} contains the estimate of
> -variant effect on the corresponding compound phenotype (\code{beta_c}), standard error (\code{s.e.}),
> +The function returns a list of 3 matrices. \code{$replication} contains the estimate of 
> +variant effect on the corresponding compound phenotype (\code{beta_c}), standard error (\code{s.e.}), 
> replication P-value (\code{P}), and proportion of phenotypic variance explained (\code{R-squared}).
> -\code{$training.coef} contains the estimated coefficients in the discovery sample of each phenotype
> -for each variant to construct the compound phenotype. \code{$test.coef} contains similar coefficients
> +\code{$training.coef} contains the estimated coefficients in the discovery sample of each phenotype 
> +for each variant to construct the compound phenotype. \code{$test.coef} contains similar coefficients 
> as in \code{$training.coef} but estimated in the replication sample, but these are just for the record,
> NOT used in the replication procedure.
> }
> @@ -46,7 +46,7 @@
> }
> \note{
> Either \code{.pheno} or \code{.phenofile} has to be provided.
> -If both are provided, only \code{phenofile} will be used. Individual IDs
> +If both are provided, only \code{phenofile} will be used. Individual IDs 
> in \code{.pheno} or \code{.phenofile} and \code{.geno} have to match!
> }
> \examples{
> @@ -55,7 +55,7 @@
> data(ge03d2)
> 
> ## running multivariate GWAS for 3 traits: height, weight, bmi
> -res <- Multivariate(gwaa.data = ge03d2, trait.cols = c(5, 6, 8),
> +res <- Multivariate(gwaa.data = ge03d2, trait.cols = c(5, 6, 8), 
>                     covariate.cols = c(2, 3))
> 
> ## extracting 5 significant variants
> @@ -70,7 +70,7 @@
> test.geno <- as.double(gtdata(ge03d2c)[,snps])
> 
> ## try replication
> -rep <- MultiRep(training.pheno = phdata(ge03d2), test.pheno = phdata(ge03d2c),
> +rep <- MultiRep(training.pheno = phdata(ge03d2), test.pheno = phdata(ge03d2c), 
>                 pheno.names = c('height', 'weight', 'bmi'),
>                    training.geno = training.geno, test.geno = test.geno)
> }
> @@ -80,7 +80,7 @@
> }
> \references{
> Xia Shen, ..., Gordan Lauc, Jim Wilson, Yurii Aulchenko (2014).
> -Multi-omic-variate analysis identified the association between 14q32.33 and
> +Multi-omic-variate analysis identified the association between 14q32.33 and 
> compound N-Glycosylation of human Immunoglobulin G \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/man/MultiSummary.Rd
> ===================================================================
> --- pkg/MultiABEL/man/MultiSummary.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/MultiSummary.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/MultiSummary.R
> \name{MultiSummary}
> \alias{MultiSummary}
> @@ -6,18 +6,22 @@
> \alias{multi.summary}
> \title{Multivariate genome-wide association scan using summary statistics}
> \usage{
> -MultiSummary(x, type = "outbred", vars = NULL)
> +MultiSummary(x, index = NULL, type = "outbred", vars = NULL)
> }
> \arguments{
> \item{x}{A data object of class \code{multi.summary} loaded by the function \code{load.summary}.}
> 
> -\item{type}{A string gives the type of analysis. Default is \code{"outbred"}, referring to
> -general outbred populations, following Hardy-Weinberg equilibrium. \code{"inbred"} refers to
> +\item{index}{A numeric vector that gives the indices of the traits to be analyzed jointly.}
> +
> +\item{type}{A string gives the type of analysis. Default is \code{"outbred"}, referring to 
> +general outbred populations, following Hardy-Weinberg equilibrium. \code{"inbred"} refers to 
> inbred populations, where no heterzygotes exists, namely, allele frequency = genotype frequency.
> -\code{"precise"} refers to precise test statistics, especially when the individual-level data
> -are available, for which the argument \code{vars} has to be given.}
> +\code{"precise"} refers to precise test statistics, especially when the individual-level data 
> +are available, for which the argument \code{vars} has to be given. \code{"direct"} refers to 
> +test statistics directly constructed from the T-statistics in univariate GWAS, this provides a
> +scale-invariant test most similar to the direct MANOVA, but may be less powerful in some scenarios.}
> 
> -\item{vars}{A numeric vector gives the variance of the genotypes at each SNP, coded as 0, 1 and 2.
> +\item{vars}{A numeric vector gives the variance of the genotypes at each SNP, e.g. coded as 0, 1 and 2.
> Only used when \code{type = "precise"}.}
> }
> \value{
> @@ -45,8 +49,8 @@
> indep.snps <- as.character(read.table('indep.snps')$V1)
> 
> ## load summary statistics of the six traits
> -stats.male <- load.summary(files = c('bmi.txt', 'height.txt',
> -                           'weight.txt', 'hip.txt', 'wc.txt',
> +stats.male <- load.summary(files = c('bmi.txt', 'height.txt', 
> +                           'weight.txt', 'hip.txt', 'wc.txt', 
>                            'whr.txt'), indep.snps = indep.snps)
> 
> ## perform multi-trait meta-GWAS
> @@ -58,10 +62,10 @@
> Xia Shen
> }
> \references{
> -Xia Shen, Xiao Wang, Zheng Ning, Yakov Tsepilov, Masoud Shirali,
> -Generation Scotland, Blair H. Smith, Lynne J. Hocking, Sandosh Padmanabhan, Caroline Hayward,
> +Xia Shen, Zheng Ning, Yakov Tsepilov, Masoud Shirali,  
> +Generation Scotland, Blair H. Smith, Lynne J. Hocking, Sandosh Padmanabhan, Caroline Hayward, 
> David J. Porteous, Yudi Pawitan, Chris S. Haley, Yurii S. Aulchenko (2015).
> -Simple multi-trait analysis identifies novel loci
> +Simple multi-trait analysis identifies novel loci 
> associated with growth and obesity measures. \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/man/Multivariate.Rd
> ===================================================================
> --- pkg/MultiABEL/man/Multivariate.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/Multivariate.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,4 +1,4 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/Multivariate.R
> \name{Multivariate}
> \alias{Multivariate}
> @@ -6,11 +6,13 @@
> \alias{multivariate}
> \title{Multivariate genome-wide association scan}
> \usage{
> -Multivariate(x, ...)
> +Multivariate(x, trait.idx = NULL, ...)
> }
> \arguments{
> \item{x}{An object created by \code{\link{MultiLoad}}.}
> 
> +\item{trait.idx}{A vector giving the indices of traits to be analyzed.}
> +
> \item{...}{not used.}
> }
> \value{
> @@ -26,7 +28,7 @@
> analysis of variance (MANOVA).
> }
> \note{
> -Either \code{gwaa.data} (for GenABEL data format) or the combination of
> +Either \code{gwaa.data} (for GenABEL data format) or the combination of 
> \code{phenofile} and \code{genofile} (for DatABEL data format) has to be provided.
> If all are provided, only \code{phenofile} and \code{genofile} will be used. When using
> DatABEL format input, individual IDs in \code{phenofile} and \code{genofile} have to match!
> @@ -37,7 +39,7 @@
> data(ge03d2ex.clean)
> 
> ## running multivariate GWAS for 3 traits: height, weight, bmi
> -loaded <- Multivariate(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8),
> +loaded <- MultiLoad(gwaa.data = ge03d2ex.clean, trait.cols = c(5, 6, 8), 
>                     covariate.cols = c(2, 3))
> 
> ## running the multivariate GWAS again
> @@ -49,7 +51,7 @@
> }
> \references{
> Xia Shen, ..., Gordan Lauc, Jim Wilson, Yurii Aulchenko (2015).
> -Multi-omic-variate analysis identified novel loci associated with
> +Multi-omic-variate analysis identified novel loci associated with 
> compound N-Glycosylation of human Immunoglobulin G. \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/man/load.summary.Rd
> ===================================================================
> --- pkg/MultiABEL/man/load.summary.Rd	2016-03-23 01:12:36 UTC (rev 2051)
> +++ pkg/MultiABEL/man/load.summary.Rd	2016-04-16 20:39:42 UTC (rev 2052)
> @@ -1,35 +1,56 @@
> -% Generated by roxygen2 (4.1.1): do not edit by hand
> +% Generated by roxygen2: do not edit by hand
> % Please edit documentation in R/load.summary.R
> \name{load.summary}
> \alias{load.summary}
> \title{Loading multiple summary statistics from genome-wide association studies}
> \usage{
> -load.summary(files, cor.pheno = NULL, indep.snps = NULL)
> +load.summary(files, cor.pheno = NULL, indep.snps = NULL, est.var = FALSE,
> +  type = "outbred", vars = NULL, columnNames = c("snp", "a1", "freq",
> +  "beta", "se", "n"), fixedN = NULL)
> }
> \arguments{
> \item{files}{A vector of file names as strings. Each file name should contain summary statistics of
> one trait to be included in the multi-trait analysis. The columns of the summary statistics have to
> -contain \code{'snp'} (marker ID), \code{'a1'} (the first allele), \code{'freq'}
> -(frequency of the first allele), \code{'beta'} (effect size), \code{'se'} (standard error), and
> +contain (uppercase or lowercase does not matter) \code{'snp'} (marker ID), \code{'a1'} (the first allele), \code{'freq'} 
> +(frequency of the first allele), \code{'beta'} (effect size), \code{'se'} (standard error), and 
> \code{'n'} (sample size).}
> 
> -\item{cor.pheno}{A #traits x #traits matrix of correlation matrix of the phenotypes, to be used to
> +\item{cor.pheno}{A #traits x #traits matrix of correlation matrix of the phenotypes, to be used to 
> construct the multi-trait test statistic. If \code{NULL},
> -this matrix will be estimated from genome-wide summary statistics. If you have partially overlapping
> +this matrix will be estimated from genome-wide summary statistics. If you have partially overlapping 
> samples for different traits, shrinkage correlation matrix is recommended (see reference), so in that
> case, unless you know what you are doing, leave this argument as default, i.e. \code{NULL}.}
> 
> -\item{indep.snps}{A vector of strings containing the names of a set of independent SNPs. This is
> -recommended to be generated by LD-pruning the genotype data in a certain cohort. Typically the
> +\item{indep.snps}{A vector of strings containing the names of a set of independent SNPs. This is 
> +recommended to be generated by LD-pruning the genotype data in a certain cohort. Typically the 
> number of SNPs should be more than 10,000 in order to obtain a good estimate of \code{cor.pheno}. If
> \code{cor.pheno = NULL}, this argument cannot be \code{NULL}.}
> +
> +\item{est.var}{A logical value. If \code{FALSE}, each phenotypic variance is assumed to be known as 1. 
> +If \code{TRUE}, each phenotypic variance will be estimated to adjust the summary statistics, so that
> +the corresponding phenoypic variance is 1.}
> +
> +\item{type}{A string gives the type of analysis. Default is \code{"outbred"}, referring to 
> +general outbred populations, following Hardy-Weinberg equilibrium. \code{"inbred"} refers to 
> +inbred populations, where no heterzygotes exists, namely, allele frequency = genotype frequency.
> +\code{"precise"} refers to precise genotypic variance, especially when the individual-level data 
> +are available, for which the argument \code{vars} has to be given.}
> +
> +\item{vars}{A numeric vector gives the variance of the genotypes at each SNP, e.g. coded as 0, 1 and 2.
> +Only used when \code{type = "precise"}.}
> +
> +\item{columnNames}{A vector with names of columns containing necessary information in the input file;
> +default values are c('snp','a1','freq','beta','se','n'). The values are case-insensitive.}
> +
> +\item{fixedN}{sample size to assume across all analyses, when provided, this number will be used
> +(instead of the ones specified in the input files)}
> }
> \value{
> The function returns a list of class \code{multi.summary}, containing two elements: \code{gwa}
> (the cleaned data to be processed in multi-trait GWAS) and \code{cor.pheno} (user input or estimated).
> }
> \description{
> -The function loads multiple meta-GWAS summary statistics, for subsequent multi-trait GWAS.
> +The function loads multiple meta-GWAS summary statistics, for subsequent multi-trait GWAS. 
> Currently, the package only analyzes summary statistics from inverse-Gaussianized continuous traits.
> }
> \examples{
> @@ -55,13 +76,13 @@
> }
> }
> \author{
> -Xia Shen
> +Xia Shen, Yurii Aulchenko
> }
> \references{
> -Xia Shen, Xiao Wang, Zheng Ning, Yakov Tsepilov, Masoud Shirali,
> -Generation Scotland, Blair H. Smith, Lynne J. Hocking, Sandosh Padmanabhan, Caroline Hayward,
> +Xia Shen, Zheng Ning, Yakov Tsepilov, Masoud Shirali, 
> +Generation Scotland, Blair H. Smith, Lynne J. Hocking, Sandosh Padmanabhan, Caroline Hayward, 
> David J. Porteous, Yudi Pawitan, Chris S. Haley, Yurii S. Aulchenko (2015).
> -Simple multi-trait analysis identifies novel loci
> +Simple multi-trait analysis identifies novel loci 
> associated with growth and obesity measures. \emph{Submitted}.
> }
> \seealso{
> 
> Modified: pkg/MultiABEL/src/symbols.rds
> ===================================================================
> (Binary files differ)
> 
> _______________________________________________
> Genabel-commits mailing list
> Genabel-commits at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-commits


From lennart at karssen.org  Sat Apr 30 15:49:10 2016
From: lennart at karssen.org (L.C. Karssen)
Date: Sat, 30 Apr 2016 15:49:10 +0200
Subject: [GenABEL-dev] Upcoming ProbABEL release and Eigen lib inclusion
Message-ID: <5724B7D6.1040804@karssen.org>

cDear list,

I'm preparing a new release of ProbABEL (probably v0.5.0), with a bunch
of bug fixes (see the Changelog [1]). One of the major changes is that
ProbABEL now fully depends on the Eigen library for matrix manipulation.

Currently, only one thing stands in the way of finalising the release:
Issue #11 [2]. Due to a change in the Eigen library starting from
version 3.2.2, some of the ProbABEL checks where we compare R output
with ProbABEL output fail. Luckily these failures are only in those
checks that test corner cases (e.g. our test SNP 6, where only 1 out of
182 samples has a different genotype).
What I observe is that with Eigen v3.2.1 the output of ProbABEL is equal
to that produced by R. With newer Eigen versions, there are differences
(although in the end, the result is still that there is no association,
e.g. small betas and huge standard errors).

In my opinion the proper solution of issue #11 would be to write proper
checks for these corner case (like singular matrices) and output
warnings if they occur. However, that would delay the release of
ProbABEL too much.

So my suggestion is to include Eigen v3.2.1 with ProbABEL v0.5.0. This
way we will be certain that ProbABEL and R give the same output, and,
addtionally, having Eigen included in the ProbABEL tar.gz will make life
easier for a (novice) user installing ProbABEL from source. Meanwhile
this will buy us time to fix issue #11.

Regarding licences, I don't see a problem. Eigen is licensed under the
Mozilla Public Licence (MPL) [3, 4], which allows distribution of
(unmodified code) in GPL software [5].

I would like to hear your opinions on distribution Eigen with ProbABEL.
Is there anything I overlooked?


Best regards,

Lennart.


[1] https://github.com/GenABEL-Project/ProbABEL/blob/master/doc/ChangeLog
[2] https://github.com/GenABEL-Project/ProbABEL/issues/11
[3] http://eigen.tuxfamily.org/index.php?title=Main_Page#License
[4] https://www.mozilla.org/en-US/MPL/2.0/
[5] https://www.mozilla.org/en-US/MPL/2.0/combining-mpl-and-gpl/,
especially section "Unmodified MPL-licensed Files - MPL-only"
-- 
*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
L.C. Karssen
Utrecht
The Netherlands

lennart at karssen.org
http://blog.karssen.org
GPG key ID: A88F554A
-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 213 bytes
Desc: OpenPGP digital signature
URL: <http://lists.r-forge.r-project.org/pipermail/genabel-devel/attachments/20160430/4a4f5628/attachment.sig>

From yurii.aulchenko at gmail.com  Sat Apr 30 18:29:12 2016
From: yurii.aulchenko at gmail.com (Yurii Aulchenko)
Date: Sat, 30 Apr 2016 18:29:12 +0200
Subject: [GenABEL-dev] Upcoming ProbABEL release and Eigen lib inclusion
In-Reply-To: <5724B7D6.1040804@karssen.org>
References: <5724B7D6.1040804@karssen.org>
Message-ID: <8F456C58-8F38-435D-B3F3-0400668CCCB2@gmail.com>

Sounds like reasonable solution to me. Yurii

----------------------
Yurii Aulchenko 
(sent from mobile device)

> On 30 Apr 2016, at 15:49, L.C. Karssen <lennart at karssen.org> wrote:
> 
> cDear list,
> 
> I'm preparing a new release of ProbABEL (probably v0.5.0), with a bunch
> of bug fixes (see the Changelog [1]). One of the major changes is that
> ProbABEL now fully depends on the Eigen library for matrix manipulation.
> 
> Currently, only one thing stands in the way of finalising the release:
> Issue #11 [2]. Due to a change in the Eigen library starting from
> version 3.2.2, some of the ProbABEL checks where we compare R output
> with ProbABEL output fail. Luckily these failures are only in those
> checks that test corner cases (e.g. our test SNP 6, where only 1 out of
> 182 samples has a different genotype).
> What I observe is that with Eigen v3.2.1 the output of ProbABEL is equal
> to that produced by R. With newer Eigen versions, there are differences
> (although in the end, the result is still that there is no association,
> e.g. small betas and huge standard errors).
> 
> In my opinion the proper solution of issue #11 would be to write proper
> checks for these corner case (like singular matrices) and output
> warnings if they occur. However, that would delay the release of
> ProbABEL too much.
> 
> So my suggestion is to include Eigen v3.2.1 with ProbABEL v0.5.0. This
> way we will be certain that ProbABEL and R give the same output, and,
> addtionally, having Eigen included in the ProbABEL tar.gz will make life
> easier for a (novice) user installing ProbABEL from source. Meanwhile
> this will buy us time to fix issue #11.
> 
> Regarding licences, I don't see a problem. Eigen is licensed under the
> Mozilla Public Licence (MPL) [3, 4], which allows distribution of
> (unmodified code) in GPL software [5].
> 
> I would like to hear your opinions on distribution Eigen with ProbABEL.
> Is there anything I overlooked?
> 
> 
> Best regards,
> 
> Lennart.
> 
> 
> [1] https://github.com/GenABEL-Project/ProbABEL/blob/master/doc/ChangeLog
> [2] https://github.com/GenABEL-Project/ProbABEL/issues/11
> [3] http://eigen.tuxfamily.org/index.php?title=Main_Page#License
> [4] https://www.mozilla.org/en-US/MPL/2.0/
> [5] https://www.mozilla.org/en-US/MPL/2.0/combining-mpl-and-gpl/,
> especially section "Unmodified MPL-licensed Files - MPL-only"
> -- 
> *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
> L.C. Karssen
> Utrecht
> The Netherlands
> 
> lennart at karssen.org
> http://blog.karssen.org
> GPG key ID: A88F554A
> -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-
> 
> _______________________________________________
> genabel-devel mailing list
> genabel-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-devel

From fabregat at aices.rwth-aachen.de  Sat Apr 30 19:09:35 2016
From: fabregat at aices.rwth-aachen.de (Diego Fabregat)
Date: Sat, 30 Apr 2016 19:09:35 +0200
Subject: [GenABEL-dev] Upcoming ProbABEL release and Eigen lib inclusion
In-Reply-To: <5724B7D6.1040804@karssen.org>
References: <5724B7D6.1040804@karssen.org>
Message-ID: <5724E6CF.10702@aices.rwth-aachen.de>



On 30.04.2016 15:49, L.C. Karssen wrote:
> Regarding licences, I don't see a problem. Eigen is licensed under the
> Mozilla Public Licence (MPL) [3, 4], which allows distribution of
> (unmodified code) in GPL software [5].
>
> I would like to hear your opinions on distribution Eigen with ProbABEL.
> Is there anything I overlooked?
According to [5], it looks like there is no problem in doing that :)

If you want that people use ProbABEL with that specific version of
Eigen, it seems to be the simplest way for the user.

> [1] https://github.com/GenABEL-Project/ProbABEL/blob/master/doc/ChangeLog
> [2] https://github.com/GenABEL-Project/ProbABEL/issues/11
> [3] http://eigen.tuxfamily.org/index.php?title=Main_Page#License
> [4] https://www.mozilla.org/en-US/MPL/2.0/
> [5] https://www.mozilla.org/en-US/MPL/2.0/combining-mpl-and-gpl/,
> especially section "Unmodified MPL-licensed Files - MPL-only"