Hi,
I'm using DESeq2 to analyze a pooled CRISPR screen where thousands of precise mutations are induced in different yeast cells (SNPs, indels), reflecting natural variation. More information on the experimental design and analysis is provided below.
Pools of edited yeast cells were subjected to diverse conditions, with each condition having biological replicates available for analysis. Subsequently, these biological replicates were monitored over time, with samples taken at various time points.
The objective of the experiment was to identify mutations that exhibited higher or lower fitness over time, considering each condition independently. In the initial analysis, I employed DESeq2 to calculate LFCs by comparing the abundance of the edited cells between the furthest time points. Size factors were calculated based on a negative control group, comprising a subset of 300 unedited yeast cells also present in the pool. By doing so, the control group, which served as the reference for comparing the edited yeast cells, was centered around zero.
Obtained LFCs were thus for contrast between the furthest time points. However, a bimodal distribution was observed for some conditions when plotting histograms of the obtained LFC. We're really not sure what could cause this bimodal distribution as there seemed no correlation between an extreme environment and a mixture model (See figure)
Hence the questions:
- Has anybody encountered similar results during an analysis of CRISPR screens or HTS data in general?
- What could be the cause of this phenomenon?
- What are the implications of this for further analysis?
Hopefully, somebody can help me out!
