decreases simply indicates the empirical observation of an apparent causal relationship under controlled experimental conditions, such as the relationship between an EGFR inhibitor and decreased Cell Cycle and E2F transcriptional activity in the context of specific cell lines.
directlyDecreases applies to the subset of those cases in which there is evidence that the observed causal relationship is mediated by direct physical interaction. For example, if it is demonstrated that a phosphatase can deactivate another class of protein by direct action - dephosphorlyation - then that is best represented by a directlyDecreases relationship. It would also be correct to say “decreases” but it would not be as specific and would fail to capture useful information.