Computational Multitarget Drug Discovery

Abstract

Pharmaceutical substances have been discovered by means ranging from serendipitous observation [1,2] to specific engineering [3]. The purpose is nearly always to combat one particular disease, and the approach is most often trial and error. The efficiency of these pharmaceutical hunts has been improved greatly by highthroughput pharmaceutical platforms, but the requirement of physical experiment makes these screens scale in expense linearly at best. The expense of discovering a new chemical entity is estimated at US$0.5B–US$2B [4,5]. More recent successes in computationalmodeling of compound to protein docking open the possibility of nonphysical prelaboratory screens. In our experience this has vastly increased the success rate of bench experiments [6,7] (see Table 13.1, later in this chapter). Computational modeling of protein ligand interactions has been applied to find pharmacologic targets in known drug-disease pairs [8,9]. Themore obvious use of these dockingmethods is to guide discovery of a drug for a disease, asmodeling enables design [3]. Design does not need to be limited to one protein target. Searching for one compound for multiple targets in the same pathogen increases odds for successful inhibition of at least one target, and facilitates discovery of multitarget lead inhibitors, [Note 1], which vastly decreases the probability of developing resistance (or habituation) and decreases toxicity via lowered effective dose [6,10–12] (Table 13.1). Thus far the search for multitarget inhibitors has focused on one organism at a time [6,9,11], but modeling multidisease effects has explained clinical patterns of elimination for twodiseasesbyonedrug [13].Theadvent of computationalmultidisease screens will enable access to the most accurate aspects of computational screening, bearing the possibility of vastly reducing barriers to drug development.