Uneven number of biological reps on DESeq

[sulfatereducers]

Hello.
I am relatively new to RNASeq. I just finished a PCA analysis on my libraries and on of my treatments is clustering with the controls. Worst case scenario I decide I cannot use that library for my analysis, which leaves me comparing 3 vs 2 libraries. I just want to confirm/hear comments on how valid this is. Any rationale for either response?

[MatthewHaas]

Uneven number of biological reps on DEseq.

You should have at least three replicates for each sample. I had something similar happen to me with the Genomics Center sequencing one of my samples twice and one sample not at all...In any case, I ultimately re-submitted the missing sample. The PCA plots and differential expression results did not change in any meaningful way. Best of luck.

[gringer]

I'd suggest looking for contamination in that sample (e.g. using BLAST to find out what the most abundant read maps to).

At least three replicates should be used to determine which (if any) samples are outliers. If you've found an outlier, and have another reason to exclude it, then it would be reasonable to exclude that sample from the analysis. Ideally, you would do another library prep and sequencing to make up the numbers, but people tend to dislike spending more money for statistical robustness.