Fernando, Capitalize the "b" in the "Batch" header, and let me know if this works. Alternatively, use the ComBat from the sva package of bioconductor. This will be less finicky. Evan On Jun 4, 2013, at 11:12 AM, Fernando Andrade wrote: > Hi there, > > This is the first time using combat, and i'm using it on a very large set of data built up in 14 batches. I'm gettin many error messages when running it, and i've searched for every single one of them in ComBat forum and tried to fix using the information available in the forum, but i'm still getting errors. Every time that I use tab- separeted values i get an error saying that a I have non numerical values, then i move to .csv ant the error changes: > > Reading Sample Information File > Reading Expression Data File > Found 0 batches > Found 0 covariate(s) > Standardizing Data across genes > Error in solve.default(t(design) %*% design) : > Lapack routine dgesv: system is exactly singular: U[1,1] = 0 > > or > > Reading Sample Information File > Reading Expression Data File > Found 0 batches > Found 5 covariate(s) > Error in matrix(0, length(vec), nlevels(vec) - start + 1) : > invalid 'ncol' value (< 0) > > my expression file is formatted like this: > > probe lineage1 lineage2 lineage3... > xxx 11.111 22.2222 333.3333 .... > ... > > and my sif is like this: > > ID batch > lineage1 1 > lineage2 1 > lineage3 2 > lineage4 3... > > I'd be very grateful for any help! > thanks for your attention > > Fernando

Hi there, This is the first time using combat, and i'm using it on a very large set of data built up in 14 batches. I'm gettin many error messages when running it, and i've searched for every single one of them in ComBat forum and tried to fix using the information available in the forum, but i'm still getting errors. Every time that I use tab-separeted values i get an error saying that a I have non numerical values, then i move to .csv ant the error changes: Reading Sample Information File Reading Expression Data File Found 0 batches Found 0 covariate(s) Standardizing Data across genes Error in solve.default(t(design) %*% design) : Lapack routine dgesv: system is exactly singular: U[1,1] = 0 or Reading Sample Information File Reading Expression Data File Found 0 batches Found 5 covariate(s) Error in matrix(0, length(vec), nlevels(vec) - start + 1) : invalid 'ncol' value (< 0) my expression file is formatted like this: probe lineage1 lineage2 lineage3... xxx 11.111 22.2222 333.3333 .... ... and my sif is like this: ID batch lineage1 1 lineage2 1 lineage3 2 lineage4 3... I'd be very grateful for any help! thanks for your attention Fernando [[alternative HTML version deleted]]

Fernando, Do you have any covariates to include? If so, suppose it was in the "mycov" variable as below: > mycov = c("Control", "Control", "Exp1", "Exp1", "Exp2", "Exp2") Then make the matrix: > mod = model.matrix(~mycov) > mod (Intercept) mycovExp1 mycovExp2 1 1 0 0 2 1 0 0 3 1 1 0 4 1 1 0 5 1 0 1 6 1 0 1 attr(,"assign") [1] 0 1 1 attr(,"contrasts") attr(,"contrasts")$mycov [1] "contr.treatment" Then apply ComBat: > ComBat(mydata, batch, mod) If you have no covariates to include, then just pass a "NULL" object: > ComBat(mydata, batch, NULL) Hope this helps! Evan On Jun 4, 2013, at 12:45 PM, Fernando Andrade wrote: Sorry to bother again... is there any exemple file where I can find how to make matrix models.. a exemple usage of combat? just for me to use it as a model for my data Many thanks Fernando 2013/6/4 Fernando Andrade <fernand04ndrade@gmail.com<mailto:fernand04ndrade@gmail.com>> my mistake, the 'b' is already captalized in my original sif. Sorry about that. i've just installed package sva, but i could not figure what model matrix is all about. I only have expression file and a sif. i do not have other covariates besides batch. 2013/6/4 Johnson, William Evan <wej@bu.edu<mailto:wej@bu.edu>> Fernando, Capitalize the "b" in the "Batch" header, and let me know if this works. Alternatively, use the ComBat from the sva package of bioconductor. This will be less finicky. Evan On Jun 4, 2013, at 11:12 AM, Fernando Andrade wrote: > Hi there, > > This is the first time using combat, and i'm using it on a very large set of data built up in 14 batches. I'm gettin many error messages when running it, and i've searched for every single one of them in ComBat forum and tried to fix using the information available in the forum, but i'm still getting errors. Every time that I use tab- separeted values i get an error saying that a I have non numerical values, then i move to .csv ant the error changes: > > Reading Sample Information File > Reading Expression Data File > Found 0 batches > Found 0 covariate(s) > Standardizing Data across genes > Error in solve.default(t(design) %*% design) : > Lapack routine dgesv: system is exactly singular: U[1,1] = 0 > > or > > Reading Sample Information File > Reading Expression Data File > Found 0 batches > Found 5 covariate(s) > Error in matrix(0, length(vec), nlevels(vec) - start + 1) : > invalid 'ncol' value (< 0) > > my expression file is formatted like this: > > probe lineage1 lineage2 lineage3... > xxx 11.111 22.2222 333.3333 .... > ... > > and my sif is like this: > > ID batch > lineage1 1 > lineage2 1 > lineage3 2 > lineage4 3... > > I'd be very grateful for any help! > thanks for your attention > > Fernando [[alternative HTML version deleted]]