Hello everybody,
After finding quite a lot of help here with my last issue, I have returned once more, desperate.
Thanks to the posters on my previous post, I was able to load and execute ComBat-Seq for removal of batch effects on 2 seperately sequenced data-sets.
Sadly I have run into a new issue. My samples belong to different group dependent on their timepoint+genotype and batch 1 or 2 dependent on their sequencing date.
Most of my batches work fine with ComBat, but some batches in the group only contains 1 sample, which ComBat dislikes very much. To circumvent this, and please have mercy on me as I was desperate, I have created artificial dummy samples as replicates of the single sample (i.e. X_v2) in the batch to run ComBat, and the modified its output to remove said dummy sample before passing it for further modulation to DESeq2.
As many others these samples resulted in ridiculously high integer values when other samples values were 0 or close to it, so I by-passed this by setting those as max integer values instead. I was happy, the script ran fine, up until I reached a comparison with 2 groups, whose batch 2 contained each only a single sample.
My problem: ComBat loads and runs fine, up until it is adjusting the data. It runs and runs, and has been, for similiar sample and gene size as the other groups (~1 min duration) for over 19h already.
I am at my wits end. Does anybody experienced this before or could offer a solution? I read through documentation and everything, and while I know that the issue lies with the single sample batches in both groups, I have not a single clue how to even approach this issue as I never get an actual crash or error
Thanks in advance
Code:
txi <- tximport(files, type = "salmon", ignoreTxVersion=TRUE, tx2gene = tx2gene)
combat_batch <- samples_subset$run
group_combat <- samples_subset$group
txi_subset_adjusted <- sva::ComBat_seq(txi$counts ,batch=combat_batch,group=group_combat)
samples_subset<-subset(samples_subset,first == T)
int_max = 2147483647
txi_subset_adjusted <- as.matrix(dplyr::select(as.data.frame(txi_subset_adjusted), -dplyr::contains("v2")))
txi_subset_adjusted <- ifelse(txi_subset_adjusted > int_max, int_max, txi_subset_adjusted)
ddsTxi <- DESeqDataSetFromMatrix(countData = round(txi_subset_adjusted),
colData = samples_subset,
design = ~ group)
Same keeps happening to me; but it is weird, I corrected the batch effect with the same exact function a few months ago and worked fine. I wanted to try and remove some of my batches and redo the analysis, but now the function will get stuck for ever. I have some samples where the group+batch condition only has 1 sample, but it also happened before I removed these batches. It is driving me crazy.