Dear Community,
I performed a surrogate variable analysis on my RNA-seq data using the SVA package in R and want to use the identified surrogate variables as covariates in my DE design formula. There were more than 5 significant SV identified - human samples from different locations and different timepoints, hence high heterogeneity and several confounding factors. The output of sva is anyway limited to the first 5 SV identified and I was unable to successfully increase the number of sv included in the svaseq output file. I am using the code given below:
n.sv
9
svseq=svaseq(edata,mod,mod0,n.sv=n.sv,method="two-step")
I couldn't find any helps/options adapting the output in the vignette/R help/forum, hence the question. Thanks in advance for any help!
•
written
by
robertt_bw •
Hey, Can you supply the entire code you used for this unsupervised batch effect correction?
Here you go:
edata=data.matrix(x, rownames.force = TRUE)
meta2=meta$condition
mod = model.matrix(~as.factor(meta2), data=meta2)
mod0 = model.matrix(~1, data=meta2)
n.sv = num.sv(edata,mod,method="leek")
n.sv
[1] 9
svseq=svaseq(edata,mod,mod0,n.sv=n.sv,method="two-step")
Do you spot any obvious mistake?
Seems fine, recently they developed batchQC library, here is an example.
nbatch <- 3
ncond <- 2
npercond <- 10
data.matrix <- rnaseq_sim(ngenes=50, nbatch=nbatch, ncond=ncond, npercond=
npercond, basemean=10000, ggstep=50, bbstep=2000, ccstep=800,
basedisp=100, bdispstep=-10, swvar=1000, seed=1234)
batch <- rep(1:nbatch, each=ncond*npercond)
condition <- rep(rep(1:ncond, each=npercond), nbatch)
pdata <- data.frame(batch, condition)
modmatrix = model.matrix(~as.factor(condition), data=pdata)
sva.object <- batchQC_sva(data.matrix, mod=modmatrix)
batchQC_fsva_adjusted(data.matrix, modmatrix, sva.object)
tell me what you got
Log in to answer this question.
