We suggest that in most cases, researchers should default to using MAGeCK RRA for the analysis. Our simulations suggest that it is robust and performs well in all cases.
When the screen is expected to have high variable guide efficiency, such as in CRISPRi or CRISPRa screens that investigate complex phenotypes, RRA has difficulty in finding hit genes, sometimes returning no hit genes in practice. In this case, we suggest CRISPhieRmix, as it is the only method that takes this issue into account. Although, its dependence on good controls is limiting and should be checked before use.
To my understanding, Stanley Qi lab at Stanford have developed "CRISPhieRmix" R package, especially for this issue. Inputs for running this package are the log2 fold changes for both genes and negCtrl (i.e. rho scores from the ScreenProcessing pipeline).
Storey, J. D., & Tibshirani, R. (2003). Statistical significance for genomewide studies. Proceedings of the National Academy of Sciences of the USA | Paper | GitHub
abearab
commented
1 year ago
"A benchmark of algorithms for the analysis of pooled CRISPR screens". Genome Biol (2020)
Calculating adjusted p-values in Python | stack overflow