Likelihood of side effects depends on desired clinical impact: Affinities within a very small set of targets enables inference of promiscuity or specificity of kinase inhibitors

As the heterogeneous nature of cancer starts to emerge, the focus of molecular therapy is shifting progressively towards multi-target drugs. For example, drug-based interference with several signaling pathways controlling different aspects of cell fate provides a multi-pronged attack that is proving more effective than magic bullets in hampering development and progression of malignancy. Such therapeutic agents typically target kinases, the basic signal transducers of the cell. Because kinases share common evolutionary backgrounds, they also share structural attributes, making it difficult for drugs to tell apart paralogs of clinical importance from off-target kinases. Thus, multi-target kinase inhibitors (KIs) tend to have undesired cross-reactivities with potentially lethal or debilitating side effects. As multi-target therapies are favored, a pressing issue takes the stakes: which type of clinical impact can only be achieved with a promiscuous drug, and conversely, which clinical effect lends itself to drug specificity? Combining statistical analysis with data mining and machine learning, we determine extremely small inferential bases with 3-5 targets that enable a kinomewide assessment of promiscuity and specificity with over 97% accuracy. Thus, the likelihood of side effects in molecular therapy arising from undesired cross-activities is pivotally dependent on the intended clinical impact restricted to checking a few relevant targets.