I tried to find the dependency in a specific genetic background. So, I selected two groups of cancer cell lines ,a control group and a genetic variant group, to find the dependent genes in the genetic variant group using the tools “Data Explorer” according to DepMap Public 20Q4 v2 CRISPR screening. the results are shown in a volcano plot . However, the results are confused to me. how to interpret the value of “Effect Size” in the X axis. the value, “Effect Size”, of some genes is negative and the value of the other genes is positive. the negative value represent negative genes or the positive represent positive genes?
The effect size in two-class comparison is basically the difference between average response values (in your case the dependency values) between the sets of cell lines. In particular, a positive effect size of 1 implies the mean dependency value of the in-set cell lines for that gene is 1 unit larger than the average of the out-of-set ones.
Is the effect size just the average of treatment group - average of control group or it is the above score divided by standard deviation?
Apologies for the late response. As you pointed out effect size is the difference between the averages of treatment and controls (in portal’s terminology case in-group and out-group), this difference is moderated with an empirical Bayes method using the adaptive shrinkage method described in CRAN - Package ashr
Thank you so much for providing this info. I was also interested in this functionality of DepMap, and accessed the cdsrmodels package’s lin_associations function to reproduce it as I saw mentioned in another comment of yours. However, with this function I don’t see an output of effect-size. Is this also outputtable using the package’s premade functions?
Yes, you can reproduce the results using lin_associations function. I believe the output that corresponds to the effect_size is “PosteriorMean” column of the res.table, i.e.
fit = cdsrmodels::lin_associations(X,Y)
effect.size = fit[[“res.table”]]$PosteriorMean