search

jtleek / svaseq

Analysis for svaseq paper
19 stars 11 forks source link

batch effect correction of DNA methylation data using sva #5

Closed frylhc closed 2 years ago

frylhc commented 2 years ago

Hi, Jeff, I have DNA methylation data generated from whole genome bisulfite sequencing. I want to correct unknown batch effect in my data, but encountered error message Error in eigen(t(resid) %*% resid) : infinite or missing values in 'x'. The command line I used is svaseq <- svaseq(dat,mod,mod0,n.sv=1,constant=1) . I know that the svaseq function will first apply a moderated log transform on the input dataset, I am wondering whether or not the error was caused by log transform of DNA methylation levels in the input dataset? For the DNA methylation levels in the input dataset varied from 0 to 100%. Great thanks. image

frylhc commented 2 years ago

Hi, Jeff. Sorry to bother you again. I have just understand the solution offered in BioStar, and have solved the problem. The error is caused by variables in the data set that did not respond to treatment. In my case, there are DNA methylation levels that was completely the same between biological samples and between treatment and control. This might be a potential solution for those who may encounter such a problem in the future. Again, thanks a lot.

frylhc commented 2 years ago

Anyone interested in the solution please refer to the answer of Peter Langfelder in Bioconductor.https://www.biostars.org/p/198820/