Conditions incorrectly assigned to samples

[jordimartorell]

Dear developers,

I am strugling with the conditional effects in splatPop. When I run this code, with a high cde.facLoc to see clearly the difference between conditions, and with one condition assigned to one sample and other conditions to 9 samples:

sim <- splatPopSimulate(params = newSplatPopParams(nGenes = 100,
                        batchCells=10,
                        batch.size=10,
                        condition.prob=c(0.1, 0.9),
                        cde.facLoc = 10,
                        seed=1234),
            vcf=mockVCF(n.samples=10),
            method="single",
            sparsify = FALSE) 

sim <- logNormCounts(sim)
sim <- runPCA(sim)
plotPCA(sim, colour_by = "Condition")

I get the following PCA:

It can be observed that the first principal component clearly separate the two conditions, but Condition1 is not assigned to the sample at the left, but to one of the 9 samples at the right instead. Furthermore, I've made some tests and, with this labels, I find no differential expression between conditions. However, after manual "correction" of the labels, I find a very significant DE.

I tried many different parameters in the simulation, but I always get the same results. Could you please check if there is some bug with the samples assignment to conditions?

Best regards, Jordi

[lazappi]

Hmmm...possibly there is an indexing issue somewhere? @azodichr can you please look into this?

[azodichr]

Hi @jordimartorell, Thanks for bringing this issue to our attention. I have fixed the problem with conditional group assignments getting mislabeled and pushed it to @lazappi (see pull reqest "Issue_158") to be merged into the dev version.

[lazappi]

I have merged the PR and the fix should be available on release and devel soon