Phenotypic approaches in CNS drug discovery: back to the future.

In the second installment of this blog (, I concluded that allosteric approaches hold promise, but that they have not yet delivered more efficacious CNS drugs. Perhaps no surprise in light of the high hurdle that the formidable complexity and multifactorial etiology of many CNS disorders puts on finding an efficacious mono-therapy. Indeed, even if we eventually would elucidate the molecular and cellular basis for these disorders, a drug would probably have to hit several targets with the right selectivity, potency and intrinsic activity. This seems impossible with current medicinal chemistry and structural biology approaches. So let’s turn the rudder around 180° and look at ‘target-agnostic’ (aka phenotypical or non-biased) approaches.


One of the all-time greats, in drug discovery was Paul Janssen, (1926-2003). He founded the company of the same name (now Johnson & Johnson) and was involved in the discovery of 79 (!) new medicines (Lewi and Smith, 2007). At the core of the approach of “Dr. Paul” was phenotypic drug discovery. Antipsychotics, like risperidone, were discovered by behavioral tests, such as reversal of amphetamine-induced hyperactivity in rodents. This impressive track record notwithstanding, in this age and time, where molecular and cellular sciences rule, behavioral phenotypic screening can feel somewhat dated. But consider the following statistics: a review of FDA drug approvals from 1999-2008 showed that 28 out of the 50 approved, first-in-class, small molecular weight drugs were discovered through phenotypic screening. This ratio is even more impressive for CNS disorders: 6 out of 8 small drugs (Swinney and Anthony, 2011; Table 1).  Five of these compounds were discovered in animal models; of these, three were anti-epileptics discovered in (mouse) seizure models. So the more recent trackrecord of animal models is perhaps not as dire as often stated. But there is room for substantial improvement, as the value of this generation of approved CNS drugs is more incremental. Can we find better, more efficacious, drugs with more recently developed phenotypic approaches? Let’s look at a couple of examples: one behavioral and one (sub)celluar approach.

Table 1: Overview of small molecular weight CNS drugs that were discovered by phenotypic drug discovery and approved by the FDA between 1999 and 2008




How found?




Dopamine receptor

Known target and seeking improved mechanism of action

Animal models


Smoking cessation

Nicotine receptor

Known target and seeking improved mechanism of action

Neurochemistry in animals



Glutamate receptor


Not applicable




Screening random library

Animal convulsion




Screening compound specific libraries*

Animal convulsion




Screening compound specific libraries*

Animal convulsion















Source: Swinney and Anthony, 2011; * based on significant prior knowledge of compound properties


Behavioral phenotypic testing


The ‘SmartCube’ platform from Psychogenics (Tarrytown, NY) catapulted old school behavioral pharmacology into the 21st century. In a nutshell, SmartCube combines behavioral neurobiology with robotics, computer vision and bioinformatics to process and analyze massive datasets. This platform provides a sequence of challenges to a mouse, extracts more than 2000 features during a session and using proprietary bioinformatics detects the potential of compounds to treat psychiatric disorders in an unbiased way by comparing their complex behavioral profiles with those from a proprietary reference database (Alexandrov et al., 2015; Fig 1).



Figure 1: Psychogenic's Smartcube phenotypic behavioral screening platform.

The technology is highly scalable and is relatively high throughput for animal testing. Drugs originally discovered with Smartcube have recently entered clinical testing). That a SmartCube ‘signature’ can translate into the clinic was demonstrated by a positive Phase 2 trial with eltoprazine in ADHD ( Although this is certainly an encouraging finding, one cannot get overexcited yet. So far, only a few compounds were published that made it from mouse SmartCube testing to human Proof of Concept testing. And there may be at least one potential false positive finding: PDE10 inhibitors for the treatment of schizophrenia (Roberds et al., 2011).  Also, from a translational perspective, phenotypic drug discovery has unique challenges compared to target-based drug discovery. Dose selection for clinical testing is increasingly informed by PK-PD relationships, such as between plasma drug exposures and receptor occupancy. That is difficult for a phenotypical approach. Wearable devices and smart health analytics will provide a way for future medicine to provide accurate personalized diagnosis and follow up, allowing also for comprehensive and unbiased therapeutics development (something achieved in the preclinical area with the development of SmartCube, Dani Brunner, personal communincation). Also, the absence of a known molecular mechanism of action is often thought an additional hurdle for assessing and managing the risk for unwanted side effects


Cellular phenotypic testing


San Diego-based Afraxis specializes in subcellular imaging ( The company developed an enhanced dendritic spine profiling (ESP) platform for quantifying changes in synaptic connectivity using super resolution, laser-scanning, confocal microscopy. Abnormalities in spine numbers and structure are present in several CNS disorders and offer phenotypic targets for drug discovery. Like the Smartcube, ESP relies heavily on phenotypes derived from testing drugs of well-developed therapeutic classes to inform emerging profiles of novel compounds. The ESP-based in vivo high content analysis has been widely applied for phenotypic efficacy profiling of neuronal networks for drug development in neurodegenerative and psychiatric diseases. For example, the ESP platform has demonstrated correlations between effects on dendritic spines and behavioral effects of an NMDAR partial agonist (Burgdorf et al., 2015). It remains to be proven that ESP can discover clinically efficacious drugs with a novel mechanism of action; although it is still early days as ESP has been around for almost a decade shorter than, for example, the Smartcube. A potential advantage of ESP is that it may offer a more rapid, reliable, and translatable alternative for conventional animal model drug testing.




Drug discoverers are technophiles, but we have not yet figured out how to successfully use new technologies to discover drugs that are significantly better than the first generation CNS drugs from the ‘50s. Poor drug discovery performances in major areas of CNS are often attributed to the sheer complexity of the system we are attempting to understand and treat. How can we do better? SmartCube behavioral testing and the enhanced dendritic spine platform are examples of a recent generation of phenotypic technologies that utilize unbiased and highly scalable methodologies, which are designed to be permissive of large scale parallel measures – in other words, a system biology approach to drug discovery. By their nature, these are well-positioned to benefit from the coming wave of big data processing capabilities which very likely will be necessary to overcome many of the hurdles imposed by the high degree of complexity in the CNS. In a disease area where we understand so little about the pathophysiology, these methods simultaneously “build upon the shoulders of giants” in phenotypic drug discovery such as Dr. Janssen, while embracing the modern tools to address severe complexity. The largest impact will be obtained when these platforms are used together, since they cover different systems levels, including the subcellular, cellular, neuronal circuit, and behavioral level. Consistent findings across all levels will establish quality Proof of Mechanism data packages that in turn will provide a stable platform for subsequent Proof of Concept testing in higher species.




Swinney DC and Anthony J (2011) How were new medicines discovered? Nature Reviews Drug Discovery 10: 507-19


Lewi PJ and A Smith (2007) Successful pharmaceutical discovery: Paul Janssen’s concept of drug research. R&D Management 37: 355-62


Alexandrov V, Brunner D, Hanania T, Leahy E (2015) Highthroughtput analysis of behavior for drug discovery. European Journal of. Neurosci., 09 September 2011 |


Roberds SL, Filippov I, Alexandrov V, Hanania, T, Brunner D (2011). Rapid, computer vision-enabled murine screening system identifies neuropharmacological potential of two new mechanisms. Frontiers in Pharmacology 753: 127–34


Burgdorf, J., Zhang, X. -l., Weiss, C., Gross, A., Boikess, S. R., Kroes, R. A., … Moskal, J. R. (2015). The long-lasting antidepressant effects of rapastinel (GLYX-13) are associated with a metaplasticity process in the medial prefrontal cortex and hippocampus. Neuroscience 308: 202–211




I want to thank Chris Rex for the stimulating drug discovery discussions and his contribution to this blog.




Latest Comments

Blog Archive

Blog Calendar

March 2018
Mo Tu We Th Fr Sa Su
26 27 28 1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31 1