Hello Rory,

I have the following experimental design, Rep1 and Rep2 are collected in different batches, with batch effect obvious from PCA plots:

Sample    Condition    Replicate
1              A         1
2              A         2
3              B         1
4              B         2

I used the following code to account for batch effect:

db.counts <- dba.contrast(db.counts, 
                            categories=DBA_CONDITION, 
                            block = DBA_REPLICATE, # block design
                            minMember = 2)
db.counts <- dba.analyze(db.counts)
db.results <- dba.report(db.counts, th = 1, 
                         method=db.counts$config$AnalysisMethod,
                         bCalled = TRUE)

However, when I use the altered code below, all the output csv files are the same as above.

db.counts <- dba.contrast(db.counts, 
                            categories=DBA_CONDITION, 
                           # block = DBA_REPLICATE, # block design
                            minMember = 2)
db.counts <- dba.analyze(db.counts)
db.results <- dba.report(db.counts, th = 1, 
                         method=db.counts$config$AnalysisMethod,
                         bCalled = TRUE)
write.csv(db.results, paste(metaFile, "res.csv", sep = '.'))

Did I do sth. wrong? Thanks!

Ray

When you add a blocking contrast, DiffBind runs two different analyses: one with and one without the blocking factor. By default, when you call dba.report(), you get the non-blocked analysis. You have to request the blocked results using the method parameter.

To retrieve the results for the blocked analysis:

db.results <- dba.report(db.counts, th = 1, 
                         method=DBA_DESEQ2_BLOCK,
                         bCalled = TRUE)