Screen designs differ in sensitivity and dynamic range. Screening for gametocyte development, for example, involves flow sorting gametocyte-infected red blood cells. Even mutants that do not make gametocytes can be sorted into the gametocyte pot as asexuals co-infecting the same reticulocyte as a gametocyte. Co-infection rates thus limit the sensitivity of this screen. Complete loss of gametocytes in an ap2-g mutant therefore leads to a log2FC of -2.56 for male gametocytes, but to an apparently much larger effect (log2FC of -8.45) on male fertility. However, this apparent difference only reflects the maximal effect size each screen can reveal. This is important to consider when users filter on effect size. For the “phenotype summary” view, we scale all data to the dynamic range of each screen to obtain a more intuitive visualization.
Discrete phenotype categories (e.g., essential vs. dispensable) oversimplify biology because they rely on somewhat arbitrary cut-offs. Effect size and variance both affect categorization. In some screens, to reduce categorization errors we assign ‘no power’ where the statistical power to make a call is low. Filter on phenotype category to intersect two phenotype groups. Categories can also be overlayed on gene expression patterns of the Malaria Cell Atlas.
We also offer functionalities to map any gene list onto MCA co-expression clusters and to examine gene lists for certain enrichments.