How are the “Dependencies enriched in x lineage” lists determined?
For instance I’m interested in the list of dependencies that are enriched in Melanoma but am wondering how this list is compiled? There are some genes which appear on this list, that, when searched individually do not appear to be dependent in Melanoma based off of the CERES or DEMETER scores (example: IDH1). I’m also wondering what the test-statistic and p-values in this list correspond to e.g. what does a negative test statistic vs. a positive test statistic mean in this context?
Enrichment is measured by a t-test, so it’s simply assessing whether the mean of the lines which are in the context are significantly different than the mean of the lines which are not in that context.
I looked up IDH1 and didn’t see Melanoma listed as an enriched context
However, I wonder if you found this gene by looking up the Melanoma context and seeing it listed as enriched there?
These two places in the portal are actually reporting the same enrichments, only filtering the list differently.
So, what I think is happening is that IDH1 appears to be enriched, but only when you look in the RNAi data from Novartis. When you add in the lines that were screened at the Broad (what we label as “Combined RNAi”) that combined set doesn’t show enrichment.
Most places in the portal when we show RNAi, we’re using the combined RNAi dataset so I suspect that’s the source of the inconsistency you’re seeing.
This makes a lot of sense. Thanks for your help!