On discovering new medicines

July 14, 2011

ResearchBlogging.orgA link from writedit pointed me to a review of drugs that were approved in the US with an eye to how they were identified. Swinney and Anthony (2011) identified 259 agents that were approved by the US FDA between 1999 and 2008. They then identified 75 which were “first in class”, i.e., not just me-too drugs or new formulations of existing drugs or whatnot. There were 20 imaging agents, not further discussed, and 164 “follower” drugs.

The review also focused mostly on small molecule drugs instead of “biologics” because of an assumption that the latter are all exclusively “target based” discoveries. The main interest was in determining if the remaining small molecule drugs were discovered the smart way or the dumb way. That’s my formulation of what the authors term “target based screening” (which may include “molecular mechanism of action”) discovery and “phenotypic screening” type of discovery. As they put it:

The strengths of the target-based approach include the ability to apply molecular and chemical knowledge to investigate specific molecular hypotheses, and the ability to apply both small-molecule screening strategies (which can often be achieved using high-throughput formats) and biologic-based approaches, such as identifying monoclonal antibodies. A disadvantage of the target-based approach is that the solution to the specific molecular hypotheses may not be relevant to the disease pathogenesis or provide a sufficient therapeutic index.

A strength of the phenotypic approach is that the assays do not require prior understanding of the molecular mechanism of action (MMOA), and activity in such assays might be translated into therapeutic impact in a given disease state more effectively than in target-based assays, which are often more artificial. A disadvantage of phenotypic screening approaches is the challenge of optimizing the molecular properties of candidate drugs without the design parameters provided by prior knowledge of the MMOA.

You will note that this is related to some comments I made previously about mouse models of depression.

The authors found that 28 of the first-in-class new molecular entities (NMEs) were discovered via phenotypic screening, 17 via target based approaches and 5 via making synthetic mimics of existing natural compounds. To give you a flavor of what phenotypic screening means:

For example, the oxazolidinone antibiotics (such as linezolid) were initially discovered as inhibitors of Gram-positive bacteria but were subsequently shown to be protein synthesis inhibitors that target an early step in the binding of N-formylmethionyl-tRNA to the ribosome

and for target based approaches:

A computer-assisted drug design strategy that was based on the crystal structure of the influenza viral neuraminidase led to the identification of zanamivir

The authors even ventured to distinguish discovery approaches by disease area:

Evaluation of the discovery strategy by disease area showed that a phenotypic approach was the most successful for central nervous system disorders and infectious diseases, whereas target-based approaches were most successful in cancer, infectious diseases and metabolic diseases

Unsurprising of course, given that our state of understanding of nervous system disorders is, to most viewers, considerably less complete in comparison with some other health conditions. You would expect that if there are multiple targets or targets are essentially unknown, all you are left with are the predictive phenotypic models.

Of the follower drugs 51% were identified by target based discovery and 18% by phenotypic screening. This is perhaps slightly surprising given that in the cases of the me-too drugs, you would think target-based would be more heavily dominant. Perhaps we can think of a drug which initially looked to have property X that dominated but then in the phenotypic screening, it looked more like a property Y type of drug.

The authors take on this is that it is slightly surprising how poorly target-based discovery performed within a context of what they describe as a period in which there was a lot of effort and faith placed behind the target-based approaches. I suspect this is going to be in the eye of the beholder but I certainly agree. I can’t really go into the details but there are areas where my professional career is…affected, let us say…by the smart/dumb axis of drug discovery. It should be obvious to my longer term readers that I align most closely with animal models of various things related to health and neurobiology and so therefore you may safely conclude that I have a bias for phenotypic screening. And even in the case of the target-based discovery:

at least three hypotheses that must be correct to result in a new drug. The first hypothesis, which also applies to other discovery approaches, is that activity in the preclinical screens that are used to select a drug candidate will translate effectively into clinically meaningful activity in patients. The other two hypotheses are that the target that is selected is important in human disease and that the MMOA of drug candidates at the target in question is one that is capable of achieving the desired biological response.

Right. You still need good phenotypic models and ultimately you are going to have to pass human clinical trials. The authors further worry that this higher burden, especially knowing the MMoA is going to lead to some misses.

in the case of phenotypic-based screening approaches, assuming that a screening assay that translates effectively to human disease is available or can be identified, a potential key advantage of this approach over target-based approaches is that there is no preconceived idea of the MMOA and target hypothesis.

Ultimately I think this review argues quite effectively for an “all hands on deck” approach to drug discovery but it can’t help but come off as a strong caution to the folks that think that “smarter” (aka, “rational drug design”) is the only solution. Yes, this points the finger at Francis Collins’ big thrust for a new translational IC at the NIH but also at the BigPharma companies that seem to be shedding their traditional models-based, phenotypic discover research units as fast as they can. No matter which side you come down on, this is a great read with lots to think about for those of us who are interested in the discovery of new medicines.
Swinney, D., & Anthony, J. (2011). How were new medicines discovered? Nature Reviews Drug Discovery, 10 (7), 507-519 DOI: 10.1038/nrd3480

No Responses Yet to “On discovering new medicines”

  1. Arlenna Says:

    Eli Lilly has a very interesting program to do phenotypic drug discovery in an ‘open innovation’ model: https://pd2.lilly.com/pd2Web/

    They seem to be adapting relatively quickly, but in some different ways than other companies, to the new pharma business model of “(ph)arm it out.” They also pioneered Innocentive (http://www.innocentive.com), another ‘open innovation’ platform to get ideas and solutions from outside on a per-solution basis.

    Overall, I totally agree–as a chemist moving into biology who now works in a College of Pharmacy, I see it all around me. It is clear that combination approaches have to be taken, and that target-based strategies alone rely on too many assumptions about things we do not yet understand about living systems. BUT we can’t stop at the phenotypic level either, we really do need to dig all the way down and find those targets and how they work, otherwise we won’t have any idea about what is happening when resistance/tolerance inevitably develops via biological adaptation. Cells are fast, we are not.


  2. Pascale Says:

    Bravo. Even though we “think” we understand the mechanism of a disorder, diseases and their host’s responses are far more complex. For example, PKC inhibitors work great in models of diabetic complications; however, in humans, not so much. We have to take both approaches.

    I would also like to raise one of Carmines rules of science (Pam Carmines and I are collaborators, colleagues, and friends):
    All pharmaceutical agents lose specificity with time.
    Mostly because all agents ultimately have other actions not initially predicted.


  3. David Kroll Says:

    I found this to be an important passage:

    A small fraction of the first-in-class NMEs (5 out of 75) were developed as synthetic versions of natural substances (that were sometimes slightly modified), including the modified heparin fondaparinux, the porphyrin verteporfin, the biopterin cofactor sapropterin, the porphyrin precursor aminolevulinic acid and the acetylated homotaurine acamprosate (FIG. 1c). Additionally, in some cases, natural substances provided starting points for small molecule phenotypic screening (10 NMEs (FIG. 1a)) and target-based discovery (3 NMEs (FIG. 1b)). In total, 18 out of the 50 (36%) first-in-class small-molecule NMEs originated from natural substances. These numbers are consistent with those reported by Newman and Cragg for the percentage of all medicines derived from natural products, and the supposition that libraries that are derived from natural substances provide good chemical starting points for optimization. (emphasis mine)

    As the majority of pharma have largely eliminated their natural products drug discovery units, the viability of academic natural products chemistry and pharmacology units takes on even greater importance.

    Thanks for bringing this paper to my attention. The associated information on drug action, Ki/EC50, etc., are a great pharmacology primer. I think I’ll be assigning this to my class in the fall.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: