Box 1. Booting up the computers.


Whereas David Swinney's paper in Nature Reviews Drug Discovery suggests ways to develop and use complex assays that efficiently identify a target and decrease late-stage attrition,1 some biotechs think a better approach to increasing R&D productivity would rely on computational methods to identify entire networks of interacting proteins that drive disease.

"A key reason for late-stage failure may well be the focus of many developers on single therapeutic targets as the key drivers of disease. That approach ignores a fundamental fact of biology: most diseases are driven by a highly complex network of interacting pathways that can, in effect, compensate for the effects at any one target. Therefore it is unlikely that focusing on a single target within that network will give the strongest or safest therapeutic effect," said Malcolm Young, CEO of e-Therapeutics Ltd.

To deal with that problem, Young and colleagues at e-Therapeutics are taking a network pharmacology approach, "which uses computational methods in combination with in vitro and in vivo assays to generate a so-called molecular footprint of the ideal therapeutic for a given disease," said Young.5

In cancer, for example, "we first of all use a computational step to identify proteins that could be targeted to induce apoptosis in tumor cells. That is, from the outset we are focused not on a single target but on multiple targets or nodes that occur in a network of proapoptotic pathways. An analysis of that network then yields a molecular footprint or signature, revealing which node or nodes would be the most effective to hit to get a therapeutic effect. We then use that signature to help us identify and/or design compounds that hit all or most of those nodes simultaneously," said Young.

The company's lead compound, ETS2101, resulted from that process. The small molecule is in preclinical testing to treat metastatic cancer. ETS2101's mechanism is undisclosed.

"Of course, we used phenotypic and target-based screens as part of our identification of ETS2101, but we did so only in tandem with a rigorous computational approach that gave us insight into the network complexity underlying cancer," said Young.

While e-Therapeutics uses computational techniques to tackle the biological complexity of targets, BioLeap Inc. uses the computer to deal with the complexity associated with target-ligand binding.

"In our opinion, a key reason for failure in late-stage trials is that the binding mechanism of ligands to the target has been poorly worked out at the preclinical stage," said CEO David Pompliano. "There's often too much of a rush to identify a compound that simply binds with high affinity to the target, without accounting for subtle binding interactions that determine the residence time of a compound at its target and, ultimately, affect key clinical parameters" such as pharmacokinetics and pharmacodynamics.

To help work out those subtle binding interactions, BioLeap uses computational methods to generate small molecular weight fragments in silico that are ranked according to their ability to bind and modulate the X-ray crystal structure of a target. The best candidates are then synthesized by design chemists and further tested in phenotypic and target-based assays.

BioLeap is pursuing a variety of preclinical targets to treat multiple diseases, including cancer, infectious diseases and metabolic diseases.