Determining false negatives in a RNAi screen

Hello all,
I am currently engaged in a genome-wide RNAi screen and I see a lot of variability in terms of phenotype consistency/strength for the same clones. This is especially true for candidates that have relatively subtle phenotypes. I understand that this variability in penetrance is expected but what is the best way to deal with false negatives in such a large scale screen? What fraction typically turns out to be false negatives and how do you assess that? At least false positives are more easily dealt with!
Any inputs from experienced folks will be highly appreciated!
Thanks very much!

Also on a very similar note, what is the best approach for statistical analysis of data arising from a genome-wide RNAi screen as mentioned above? That is prior to, and apart from Gene set enrichment analysis type algorithms…
Thanks again!

As far as I am aware, there is no generally accepted best practice. It will depend on what you want to do with your candidates.
If it is to pull in the widest number to undertake network analysis, then I know some people accept 1 positive result out of 6 tests.
If it is to do functional tests and focused study on a gene, then most people restrict themselves to the most robust hits (e.g. 3 out of 4).
If you want to compare your results with another set, we have provided the entire quantitative dataset for an RNAi screen that was run in duplicate:

The image is for the result of 4x screening of clones that got through the first round and were then tested under 2 conditions.
Almost every clones show a different type of variability.

Although this doesn’t answer your question, I hope it helps.

Hi Jonathan,
Thank you very much for your inputs. It does help quite a bit - at least I have a clearer idea now about how to approach the problem!