Functional classification of drugs by properties of their pairwise interactions

Abstract

Multidrug treatments are increasingly important in medicine and for probing biological systems. Although many studies have focused on interactions between specific drugs, little is known about the system properties of a full drug interaction network. Like their genetic counterparts, two drugs may have no interaction, or they may interact synergistically or antagonistically to increase or suppress their individual effects. Here we use a sensitive bioluminescence technique to provide quantitative measurements of pairwise interactions among 21 antibiotics that affect growth rate in Escherichia coli. We find that the drug interaction network possesses a special property: it can be separated into classes of drugs such that any two classes interact either purely synergistically or purely antagonistically. These classes correspond directly to the cellular functions affected by the drugs. This network approach provides a new conceptual framework for understanding the functional mechanisms of drugs and their cellular targets and can be applied in systems intractable to mutant screening, biochemistry or microscopy.

Uncovering emergent interactions in three-way combinations of stressors

“…Those studies concluded that the vast majority of three-way interactions are additive or can be predicted from pairwise interactions. This important difference in results and conclusions is likely because neither of these previous studies performed a rescaling based on maximum entropy or other methods to set a baseline expectation for synergy or antagonism as done here or in previous pairwise studies [35,38]. Indeed, some of us recently re-analysed Wood et al's data using a few rescaling methods and did find the existence of several higher-order emergent interactions [48].…”

“…Our method for determining the type of three-way interaction follows previous work [35,38]. To calculate our rescaled three-way interaction measures, we use the median of replicate measurements for each experiment at a single antibiotic concentration.…”

“…When this measure is negative and sufficiently large in magnitude, the interaction is synergistic because the drugs are working together such that the bacterial growth rate is lower than would be expected based on the two singledrug effects and no drug interactions. The additive range defined below in terms of the rescaled measure is based on conservative values described by Yeh et al [35]. When considering pairwise combinations of X and Y with a third drug, Z, there are two more pairwise DA measures that describe X interacting with Z,…”

“…In previous work on pairwise interactions of drugs [35] and metabolic genes [38], drug interactions have often been classified using the definition of Bliss Independence (BI) [3]. BI defines drugs to be independent (or additive) when the per cent Table 1.…”

“…Such superficial three-drug synergies may be detrimental to the patient because additional drugs might be added that are in actuality not needed for increased pathogen killing efficiency. As a result, it is challenging to determine which specific drug combinations are most clinically relevant [3,5,35].…”