Schematic illustrations of prediction and evaluation procedures. (a). Each pair of proteins in P. falciparum that shares at least on GO Slim term of the cellular component annotation set is screened for all possible domain pairs. Comparing all interacting domain pairs according to their expectation value, we assume that the highest scoring domain interaction is governing the candidate interaction. (b) For each interaction candidate we calculate hypergeometric clustering coeffcients C
and co-expression correlation coeffcients r
. These parameters allow the domain independent assessment of a protein interaction by utilizing a logistic regression model.