The Digital Biologist

The pharmaceutical patent cliff: What's coming in 2013

In February of 2011, we reported on a new study that essentially blew away all of the standard estimates of the cost of bringing a new drug to market. The former estimate of a billion or two and 12 - 15 years in costs and time, were laid waste by new data from a careful analysis of the public filings of the largest pharma companies, showing that even the very best of them is now spending something close to US$4 billion in order to get a drug to market - while for the worst performers, that figure was something closer to US$12 billion.

Most recently, the pharma industry seems to be attempting something of a makeover, however the great headwind of expiring patents - the so-called "pharmaceutical patent cliff", is still a major obstacle to the sector's return to stability and growth.

In a revealing exposition of just how big this pharma patent cliff will be, the Fierce Pharma web site has compiled a list of the top 15 pharmaceutical patent expirations due to take effect in 2013, ranked by current sales. To give you a sense of the magnitude of this problem, the article reports that the patent cliff in 2013 could amount to a loss for the major pharma companies of about US$29 billion annually, with about 70% of this being lost to generics.

You can see the full article and the list of the top 15 pharmaceutical patent expirations of 2013 here.

© The Digital Biologist | All Rights Reserved

A new report by "The Economist" laments the lack of innovation in the biopharma sector

On "The Digital Biologist" a couple of months back, I published an article about a troubling trend in the pharmaceutical industry in which R&D investment costs are soaring even as the approval rate for new drugs actually declines. One of the reasons I posited for the apparent stagnation in this sector, was the inability of corporate leaders to abandon old practices and embrace innovation and new methodologies. "Do what you've always done and you'll get what you've always gotten".

Now a new report from "The Economist" (sponsored by Quintiles) provides some real numbers to back up this idea - and while it may not be a strictly scientific or comprehensive survey of the sector, it does provide real evidence of a kind of corporate inertia at the highest levels, with regard to embracing new ideas and changing course from the rocky road that the industry currently finds itself on.

I would encourage you to read this interesting and well written report in full, but here are some of its major findings in summary ...

Less than half of industry executives even have faith that their own R&D programs currently in place, are adequate to meet the needs of their companies.

As much of 70% of the global R&D budget for the sector may actually be wasted in fruitless programs.

Barely half of all companies who responded to the survey say that they are prioritizing change and innovation and this includes the companies who admit that the programs they do have in place are largely inadequate.

An inflexible corporate culture that fears change is cited as the leading impediment to improved innovation.

In a sector that is seen as being driven more by fear than by ambition, the companies that are doing well are those in which life science innovators are able to create a culture that recognizes and rewards effort rather than penalizing failure.

You can download a full copy of this report here

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

Papers, Scissors, Rock: A bacterial strategy game

As the old saying goes: "There's nothing new under the Sun" and indeed it turns out that Nature also invented "Paper, Scissors, Rock" many millennia before the first kids ever settled a schoolyard dispute using this game. In a recent PNAS article, researchers from the University of Washington, Seattle describe the survival dynamics of a community of three strains of bacteria that compete for resources.

In addition to a healthy, wild type strain, there is a second strain that produces the bacterial toxin colicin as well as a bunch of other proteins that allow it to survive its own toxic strategy (without which it would be shooting itself in the proverbial foot). A third strain has evolved resistance to the colicin protein produced by the second strain, by acquiring mutations that prevent colicin from binding to its own proteins.

However, in keeping with the maxim that there's no such thing as a free lunch, these latter two strains of bacteria pay a price for their competitive strategies. There is a significant cost in energy and resources for the bacteria that produces colicin and all of the accompanying proteins that enable it to survive its own offensive strategy. The colicin-resistant strain also pays a price insofar as the resistance mutations it needs to survive its competitor's colicin strategy, compromise to some degree, the functions of certain vital proteins that it also needs for its general well-being.

The scenario we have then is that the colicin producing strain can outperform the wild type bacteria in the absence of the colicin resistant strain. The colicin resistant strain can outperform the colicin producing strain when the wild type strain is not around. The wild type strain can outperform the colicin resistant strain when the colicin producing strain is not around. ... the bacterial equivalent of the "Paper, Scissors, Rock" game.

So what happens when all three strains are present at once? A very interesting dynamic of "survival of the weakest" emerges.

If one of the strains really starts to dominate the strain it grows better than, this actually favors its other competitor. If the toxin producing strain starts to dominate the wild type strain, this favors the growth of the toxin resistant strain. If the toxin resistant strain starts to dominate the toxin producing strain, this favors the growth of the wild type strain. If the wild type strain starts to dominate the toxin resistant strain, this favors the growth of the toxin producing strain.

"Paper, Scissors, Rock"

When the authors of the study cultured these three strains of bacteria in their experiments, this scenario is indeed what they observed in the laboratory. When all three strains were present in the same culture, the researchers consistently observed over many experiments, that no single strain emerges on top - the result of an interesting competitive scenario that actually encourages restraint amongst the players. The authors point out that this kind of competitive dynamic has also driven the evolution of behavioral traits in the natural world, such as mating.

While not exactly the same situation, this scenario kind of reminds me of that famous scene in the movie "A Beautiful Mind" in which John Nash, the great luminary in the field of game theory, has an epiphany in which he imagines himself and his colleagues, all competing for the attentions of the same pretty blonde. Sometimes it might be very much in everybody's best interests if disputes could be settled over a game of "Paper, Scissors, Rock" instead of a barroom brawl, a showdown with law enforcement or even a war.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

Life science beyond the spreadsheet

In this era when the average cell phone has more computing power than was available to an astronaut on one of the Apollo missions, the days when a scientist used a slide rule for numerical calculations are well behind us. In spite of this however, the great majority of life scientists working in academia or in the typical biotech/pharma laboratory, use little more than its modern equivalent, the electronic calculator or the spreadsheet, for handling their research data. This is not to say that these aren't fine tools for what they do, but when you consider the amount of computing power available to a researcher in a life science laboratory, it does seem akin to someone who works at a sawmill, still cutting planks with a hand saw.

At the other end of the spectrum from these tools for doing very general calculations, is the kind of specialized enterprise software that is either purchased or comes bundled with laboratory instruments for managing and analyzing the data that they produce. Such software generally offers a suite of the kinds of analysis and visualization that are typically required for the data that it handles. For the most part however, these packages are relatively hard-wired and inflexible. If the scientist needs for example, their data presented in a slightly different format that better matches the specific needs of their research project, they are probably out of luck since they generally don't have:

1. the source code that would enable them to modify the software,  
2. the programming skills to do it even if they had the source code,
3. the person who wrote the code to explain to them how it works,  
4. the time to do it
5. all of the above

Anyone who has worked with such software has experienced the frustrations of this scenario - "No, just show me which of these 10,000 sequences has a variant of my gene in it and does not have the HindIII restriction site". The specialized software tools that are available over the internet such as the BLAST tools for searching, retrieving and aligning biological sequence data, may relieve researchers and their IT departments from the burden of having to install and maintain the software locally, but for the most part, they are similarly inflexible in the sense that they perform a limited set of specific "one size fits all" functions.

All of this software is great if it happens to do exactly what you need it to do and in exactly the way that you want it done. It is however impossible for the scientific software writers to anticipate all of the ways in which scientists will want to use their software and this is precisely where the possibility to "roll your own" can be so invaluable. Using a computer exclusively with the software that others have written for it, may be the norm for most of the computer-connected world, but when you think about it, it's kind of like having a car that can only be driven on a very limited number of roads. The ability to define your specific task to the computer and have it process the task for you, unleashes the computer's real potential and opens up a world of possibilities that are simply not available to you if you rely exclusively on pre-packaged software.

So how does this impact the life scientist?

I think it's fair to say that there is something of a disconnect between the average life scientist's exposure to computer programming and the degree to which even a modicum of experience and training in it could benefit their research. Spreadsheets can be very powerful and made to do quite complex tasks, but they work best for the kind of numerical data that can be easily and logically ordered into a tabular format, and that can be read from and written to files that also reflect this format. I have worked in pharma companies where the Microsoft Excel spreadsheet application is pretty much the only general computing tool used by its researchers - and at almost all career levels from the most junior lab technician to the most senior executive. Furthermore, aside from the obvious examples of researchers routinely performing the same calculations over and over by hand using a calculator, I have seen many examples of problems with the handling, analysis or presentation of laboratory data for which even a simple computer script of a few lines in length, could save the researcher hours of effort.

Then what are the alternatives to this situation?

There exists a plethora of useful programming platforms of one sort or another available to the life scientist, many of them at no cost. Between spreadsheets and full-blown programming languages are the mathematical platforms for data analysis and visualization of which the best known commercial examples are MATLAB and Mathematica. These platforms offer the flexibility of a mathematical scripting language within an environment that is tailor-made for numerical data analysis and visualization. Free and open source alternatives include GNU Octave and the R Language for statistical computing.

The high-level, non-domain specific programming languages are perhaps the most flexible solution of all, insofar as they give the user the freedom to express their problem in an almost infinite variety of idioms, and most of them also include extensive mathematical and scientific libraries that remove the need to re-invent the wheel for the more common scientific computing tasks such as fitting an equation to experimental data or handling vectors and matrices. Some languages such as Python and Ruby lend themselves extremely well to the rapid scripting of solutions and sweeter still, they are available at no cost. Just open an editor, type a few lines of code, hit run and your task is done (give or take the odd bug or typo in the code). This is not to say however that these languages are not also great for creating large complex applications to solve big problems. They are. I often call Python the "swiss army kinife of computing tools" because it is so versatile and indeed, many of the world's leading technology-driven organizations such as Google, Industrial Light and Magic and NASA use it to solve their own large and complex problems. 

The Java programming language (now owned and administered by Oracle) is also widely used in the life sciences and although it is more suited to programming larger applications than writing quick scripts, it does offer the advantage of being significantly faster than Python or Ruby for more computationally intensive applications. The same Java code can conveniently be run on any computer platform that suports the Java environment (and most do) and while the rich and extensive Java programming tools are available at no cost, there is a somewhat steeper learning curve for the Java language, particularly for those with no programming experience.

Given the quantity and quality of programming and scripting tools that are available, one might wonder why they are not a part of every life scientist's arsenal, yet they are not. To be clear, there is a minority of life scientists for whom computer programming is an integral part of their daily research, but at the time of writing, they are still very much a minority and not really the intended audience of this article. It has been my experience over the couple of decades that I have worked in the life science field since leaving graduate school, that the majority of life scientists who work in laboratories in academia and industry could benefit significantly from more training in computer programming as part of their studies or onging work experience, yet most have little or no exposure to it.

Rightly or wrongly, these scientists may feel that programming is a skill for which their education and training has not prepared them and one that is time-consuming to acquire. This sentiment is easy to understand given the minimal amount of computer training in most life science curricula, the incredible breadth and depth of scientific skills that a life scientist must already master to become an effective researcher and of course, the endless pressure to publish and be productive in a fiercely competitive environment that leaves little time for excursions beyond the intellectual focus on the research at hand.

It is this author's opinion that more emphasis on computational skills, both in the life science curriculum at college and in the career training of life scientists, would benefit not only the individual researcher, but also the academic field and the life science industries for which it is the foundation. Furthermore, in keeping with my belief that computational approaches will become increasingly mainstream in biological research as they are in other fields, I see this problem becoming even more pressing in the future.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

 © The Digital Biologist | All Rights Reserved

Getting the green light for metastasis

In their habitual zeal for the next hot story, the press have done something of a disservice to the important research of a couple of scientists at the University of East Anglia in the UK. The work of Soond and Chantry that appears in the January 24th issue of Oncogene is an important step forward in our understanding of the path that a cell takes from normality to neoplasia. The preliminary nature of these published findings however, did not inhibit some in the popular press to claim that it heralded an imminent cure for many types of cancer, based upon the notion that this research has succeeded in isolating "the rogue gene" that is responsible for the spread of cancer in the body. Having heard this kind of hype from the press before (remember how in 2000, the Human Genome Project was being predicted to yield a cure for most cancers within two years?) it is easy for such fanfare to have the opposite effect and to foster cynicism. Somewhere between fanfare and cynicism however, lies the kind of cautious optimism with which most researchers in the field would view these findings and this would seem to be an appropriate response to them. What the authors describe is indeed a fundamental pathway by which cells make the transition from the kind of neighborly and cooperative way of life that is necessary to maintain the integrity of tissues and organs, to the kind of every-man-for-himself maverick cell that wants to blaze its own trail, regardless of the consequences to the organism that it was hitherto a working part of.

Through their communications with their surroundings and with neighboring cells, most cells in the body have a pretty good sense of where they are at all times, and will continue to do their thing so long as these signals are all present. Remove such a cell from its normal context however and it will invariably pursue a new course of action, often involving its own self destruction through apoptosis. For example, the signaling pathways involving cell-to-cell adhesion are generally inhibitory to apoptosis when their receptor ligands (furnished by contacts with neighboring cells) are present. Remove these contacts however, and the cell will rapidly initiate a program that leads to apoptosis. Sometimes however, it is necessary for a cell to be able to migrate from one place to another (as in wound healing for example), thereby requiring it to have (at least temporarily) the ability to exist as a free-standing cell. The carefully regulated process by which this occurs is know as the epidermal to mesenchymal transition or EMT.

So why are we talking so much about EMT?  Because EMT plays a pivotal role in the progression of many cancers. A cancer cell that can bypass the regulatory controls that maintain the epidermal state and undergo EMT, becomes a cell that is capable of metastasis and the invasion of other tissues - the hallmark of a malignant tumor cell.

Soond and Chantry focused their research on a family of ubiquitin ligase proteins that modulate the cellular machinery that degrades proteins, and which itself has far-reaching consequences for cell metabolism. In particular, they studied the protein WWP2 whose function is to tag other proteins in the cell with ubiquitin, thereby targeting them for degradation in the proteosome. WWP2 turns out to be one of those proteins that is produced as isoforms of different lengths, there being a full length isoform WWP2-FL and two shorter variants WWP2-N and WWP2-C. The authors were interested in understanding how the interactions of these various forms of WWP2 with other cellular proteins could influence the ability of a cell to migrate - a hallmark of cells that have undergone EMT.

One big "needle-in-the-haystack" question that confronted the authors was to figure out which other proteins in the cell interact with the various isoforms of WWP2. An immunoprecipitation experiment showed that WWP2 interacts with the SMAD family of proteins - a finding of particular relevance for the regulation of EMT. Under the control of TGF-β, the SMAD2 and SMAD3 proteins act in a stimultory fashion with regard to EMT, pushing the cell towards a migratory phenotype, whereas SMAD7 is currently believed to be an inhibitor of EMT. Interestingly, the authors discovered that WWP2-FL, the full length form of WWP2, interacts with all 3 SMAD proteins whereas WWP2-N interacts only with SMAD3 and WWP2-C interacts only with SMAD7. Furthermore, they found that the presence of the WWP2-N variant in the cell, increased the activity of the WWP2-FL protein.

So what was happening?

The overarching finding was that increasing the amount of WWP2 in cancer cell lines known to undergo EMT, inhibited a cell's progression towards the mesenchymal state. Increasing the amount of WWP2-FL reduced the ability of TGF-β to switch on the SMAD2 and SMAD3 genes. As a control for this experiment, they were also able to show that silencing the WWP2-FL gene using siRNA reversed this effect in the treated cells, restoring the ability of TGF-β to switch on SMAD2 and SMAD3 once more. WWP2-FL was also shown to enhance the ubiquitination of SMAD2 and SMAD3, thereby targeting them for degradation and reducing their levels in the cell. This ubiquitinating activity of WWP2-FL was further shown to be enhanced by the presence of the WWP2-N isoform. And while the presence of WWP2-FL also increased the removal rate of the EMT-inhibitory SMAD7, it was found that WWP2-C seemed to be able to act in opposition, increasing the activity of SMAD7 in the cell.

While these findings are preliminary, what they point to is an interconnected network of enhancers and inhibitors that are able to subtly modulate the equilibrium of the cell between the epidermal and mesenchymal states, based upon the relative abundances of the 3 WWP2 isoforms. This network includes interactions between these WWP2 isoforms and the SMAD family of proteins that are under the control of TGF-β, but there are also interactions between the WWP2 isoforms themselves that modify their activity. This is a system that cries out for a good computer model, although it may be necessary to do some more quantitative experiments in the lab first, in order to be able to provide a realistic set of parameters for such a model.

The system of protein interactions described in Sood and Chantry's article are an important step forward in our understanding of EMT and with further research and accumulated knowledge, a fuller understanding of them could well lead to new approaches for combating invasive cancers. Such a system is also a potential poster child for a digital biology approach to understanding the complicated equilibrium between rival cell phenotypes and perhaps even for helping us along the road to new cancer therapies.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

 

Aspirin continues to impress

Not really a digital biology story per se, but one that concerns aspirin, a drug that is dear to my heart as a result of its utterly inauspicious and low tech origins. The belief that the long term consumption of small (less than analgesic) doses of aspirin can reduce the risk for certain cancers is not new, however a scientifically rigorous demonstration of these effects had been lacking for human subjects, until now ...

This amazing small molecule, originally isolated from the bark of the willow tree which had been identified as a healing agent in folklore medicine, is  thought to have a preventive effect on certain cancers, by virtue of its impact on the NF-kappaB signaling pathway, possibly as a mediator of apoptosis. This new study published in the first 2011 issue of The Lancet, looks at the long term  (4 or more years) effects of an spirin regimen and for the first time, provides solid evidence of a significant reduction of cancer risk in humans amongst those subjects who take regular small doses of aspirin over a prolonged period. 

Here is the authors' own summary ...

"Treatment with daily aspirin for 5 years or longer reduces subsequent risk of colorectal cancer. Several lines of evidence suggest that aspirin might also reduce risk of other cancers, particularly of the gastrointestinal tract, but proof in man is lacking. We studied deaths due to cancer during and after randomised trials of daily aspirin versus control done originally for prevention of vascular events.

We used individual patient data from all randomised trials of daily aspirin versus no aspirin with mean duration of scheduled trial treatment of 4 years or longer to determine the effect of allocation to aspirin on risk of cancer death in relation to scheduled duration of trial treatment for gastrointestinal and non-gastrointestinal cancers. In three large UK trials, long-term post-trial follow-up of individual patients was obtained from death certificates and cancer registries.

In eight eligible trials (25 570 patients, 674 cancer deaths), allocation to aspirin reduced death due to cancer (pooled odds ratio [OR] 0·79, 95% CI 0·68—0·92, p=0·003). On analysis of individual patient data, which were available from seven trials (23 535 patients, 657 cancer deaths), benefit was apparent only after 5 years' follow-up (all cancers, hazard ratio [HR] 0·66, 0·50—0·87; gastrointestinal cancers, 0·46, 0·27—0·77; both p=0·003). The 20-year risk of cancer death (1634 deaths in 12 659 patients in three trials) remained lower in the aspirin groups than in the control groups (all solid cancers, HR 0·80, 0·72—0·88, p<0·0001; gastrointestinal cancers, 0·65, 0·54—0·78, p<0·0001), and benefit increased (interaction p=0·01) with scheduled duration of trial treatment (≥7·5 years: all solid cancers, 0·69, 0·54—0·88, p=0·003; gastrointestinal cancers, 0·41, 0·26—0·66, p=0·0001). The latent period before an effect on deaths was about 5 years for oesophageal, pancreatic, brain, and lung cancer, but was more delayed for stomach, colorectal, and prostate cancer. For lung and oesophageal cancer, benefit was confined to adenocarcinomas, and the overall effect on 20-year risk of cancer death was greatest for adenocarcinomas (HR 0·66, 0·56—0·77, p<0·0001). Benefit was unrelated to aspirin dose (75 mg upwards), sex, or smoking, but increased with age—the absolute reduction in 20-year risk of cancer death reaching 7·08% (2·42—11·74) at age 65 years and older.

Daily aspirin reduced deaths due to several common cancers during and after the trials. Benefit increased with duration of treatment and was consistent across the different study populations. These findings have implications for guidelines on use of aspirin and for understanding of carcinogenesis and its susceptibility to drug intervention."

Gordon Webster, the author of this review of the original Lancet article, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

Biomarkers in limbo (for now)

There has been no small amount of hand-wringing at the various the biomarker conferences over the last year or two and the predominant issue has been a familiar one to anybody who is engaged in biomarker research or drug discovery - why does the life science industry continue to see diminishing returns from ever greater investment? There is a pervasive feeling of the biomarker field having reached some kind of intellectual impasse that requires a new direction, and so it came as no surprise to me that given the opportunity to vent about the issues that concern them, this was also the dominant topic for the scientists who spoke at the various panel discussions that I attended. 

In spite of the escalation of activity and investment in the field, many of the biomarkers in current use are years or even decades old and most of them have not been substantially improved upon since their discovery. Much more useful than the current PSA test for prostate cancer for example, would be a test with which a physician could confidently triage prostate cancer patients into groups of those who would probably require medical intervention and those whose disease is relatively quiescent and who would likely die of old age before their prostate cancer ever became a real health problem.

The deluge of biological data being generated, particularly in the various ‘omics fields is undoubtedly valuable, yet it continues to accumulate at rates faster than it can be assimilated into our current understanding and its burgeoning scale confronts researchers ever more forcibly with the reality of the vast complexity of biological systems. In the biomarker field, the dominant approach of searching for correlative disease markers largely sidesteps the issue of the underlying biological complexity, and while this approach has enjoyed considerable success, the limitations of being forced to work with correlative data rather than causative knowledge are starting to become apparent.

Cancer biomarkers are a prime example of this limitation. Since we don’t yet have a good handle on the subtle chains of cause and effect that deviate a cell down the path towards neoplasia, we are forced to wait until there are obvious alarm bells ringing, signaling that something has already gone horribly wrong. The ovarian cancer marker CA125 for example, only achieves any significant prognostic accuracy once the cancer has already progressed beyond the point at which therapeutic intervention would have had a good chance of being effective. To use an analogy from the challenge of complexity in the behavioral sciences - broken glass and blood on the streets are the "markers" of a riot already in progress but what you really need for successful intervention are the early signs of unrest in the crowd before any real damage is done.

Some of the causative, systems-based  approaches to disease at the cell signaling level that were highlighted at these conferences, hold great promise for moving beyond the impasse in which the biomarker field now finds itself. The mapping of genotypic and phenotypic data onto cell signaling networks is a step in the right direction insofar as it attempts to establish an underlying causal and biological connectivity between the observed correlates, although the pathway maps that it yields are hardwired, static, representations of cellular pathways that are in reality dynamic and plastic networks. There is real cause for optimisim in the potential evolution of biomarker research away from the purely correlative approach that currently dominates the field, towards more causative approaches built upon knowledge of the behavior of the underlying biological systems.

Finally, it was telling that despite a general consensus that new approaches are needed in the biomarker field which address the issue of turning the wealth of biological data into real knowledge - the great majority of the presentations and posters at the various biomarker conferences were still focused upon advances in data gathering of one kind or another. To be fair though, it should not be expected that such a significant paradigm shift in the biomarker field will happen overnight. Real progress however, will require decision makers in the pharma and biotech industries to step outside of their comfort zones somewhat and to invest in these new approaches. Throwing ever greater sums of money at the problem using the same approaches is clearly not working and as the old adage goes - doing what you’ve always done, you’re only going to get what you’ve always gotten.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

Lilly's woes highlight the challenges of drug development

The recent announcement that Eli Lilly was halting all clinical testing of its Alzheimer's drug semagacestat, was a body blow not only to Lilly, but also to the many other companies who are developing their own Alzheimer's medications. Lilly had been developing semagacestat for more than a decade, based upon experimental work that demonstrated the compound's ability to inhibit γ-secretase, an enzyme whose activity is believed to be responsible for the formation of the amyloid plaques found in the brains of Alzheimer's patients.

Based upon the extensive tissue damage apparently caused by the formation of these amyloid plaques in the brain and the fact that their detection in the brain and spinal fluid of asymptomatic patients has been shown to be prognostic of the eventual onset of the disease, this "amyloid hypothesis" has been the cornerstone of Alzheimer's research for some time. The potential therefore, that this hypothesis is wrong, incomplete or at least fundamentally misunderstood, has sent shock-waves through a pharmaceutical industry that has at least 100 other Alzheimer's drugs currently in development, the majority of which are also predicated upon the amyloid hypothesis.

Unfortunately for Lilly, the problems for semagacestat go well beyond the drug simply not working. In its clinical trials it actually proved to be worse than the placebo, with the patients receiving the drug showing a more rapid decline in cognitive function than the patients in the placebo group. As if this was not already bad enough, there was also an increased risk of skin cancer in the patients who received semagacestat. A rash (no pun intended) of negative publicity followed the announcement of these results, amongst which were headlines proclaiming Eli Lilly's new drug "Worse than nothing". Any schadenfreude on the part of competing companies however, was probably cut short by the realization that most of them were potentially in the same boat with their own Alzheimer's programs.

A number of potential mechanistic grey areas surround the development of the current crop of Alzheimer's drugs. Firstly, the amyloid hypothesis itself is predicated upon the theory that the processing of amyloid precursor protein (APP) in the brain by γ-secretase, yields the (approximately) 40 residue β-amyloid protein whose polymerization drives the formation of the amyloid plaques - and that this process is the causative factor in the development of Alzheimer's disease. Aside from the many unanswered questions that surround this process and its role in the disease, the biological function of APP is only poorly understood. This is also further complicated by the fact that amongst other things, APP has been shown to be involved in the repair of neurons. Another potential problem for the drug developers is that in addition to APP, γ-secretase also processes other proteins such as notch, an important cell surface receptor that is involved in a plethora of developmental signaling cascades. The complex nature of the γ-secretase activity that is targeted by drugs like semagacestat, has driven some pharmaceutical companies (including Lilly) to explore the development of more "surgical" inhibitors that can block the production of amyloid protein by γ-secretase while leaving its ability to process other proteins intact. This new generation of inhibitors however, still depends upon the primacy of the amyloid hypothesis as the central process in the development of Alzheimer's disease.

It is clear that the molecular players that have already been identified as having a central role in Alzheimer's disease, are interconnected by a complex web of cause and effect that will need to be unraveled if we are ever to have a clear understanding of the etiology and development of this dreadful disease. For the time-being however, a significant portion of our knowledge of the underlying biology of Alzheimer's is still correlative and phenomenological. The oft-repeated maxim that correlation is not causation seems particularly apropos with regard to the unfortunate story of semagacestat, and as with all such misfortune, hopefully some valuable lessons can be taken away from the experience.

The damage to reputation and the loss of hundreds of millions of dollars and more than a decade in the development of semagacestat, are sobering reminders that in our attempts to intervene in these complex biological systems and to modulate their activities to our own ends, we take huge risks venturing into this challenging territory without a good map of the underlying networks of cause and effect to navigate by.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved

 

Suck it and see: Medicine by folklore

With the time and cost to bring a new drug to market now conservatively estimated at about 12 to 15 years and $800 million, it is interesting to reflect on the fact that most new drugs that reach the market are still discovered using the tried and tested "suck it and see" empirical approach with which our ancestors discovered that the bark of the Willow tree can cure headaches.

Obviously, modern drug companies do not rely on people in forests randomly chewing on the odd piece of tree bark here and there, to find the next great treatment for high cholesterol. While this approach may have enjoyed such notable successes as Aspirin - the active ingredient in the bark of the Willow and still one of the best-selling drugs ever - it required generations of random sampling reinforced by the oral tradition of folklore before its utility and efficacy became evident. Not exactly the basis on which to build a business model for a competitive pharmaceutical company.

This basic "suck it and see" approach to drug development is however still very much alive and well in modern pharmaceutical companies, although these days it is a much more deliberate and concerted process in which everything is done to stack the deck in fortune's favor. The modern high throughput screening approach is an accelerated and refined form of this empirical process, but for all of its refinement, the underlying methodology is essentially the same. Try stuff out and see what works.

The problem inherent in this approach is that you may never really understand exactly how your drug works or what other activities it might have in the patient's cells until after it has been in the clinic, at which point you are running the risk of doing actual harm to your patients. Vioxx and Avandia are household names even amongst the vast majority of the population who have never taken them, for this very reason. A significant portion of the regulatory activities involved in getting a new drug approved are actually intended to mitigate this very problem, but it is clear that even this demanding regulatory process cannot prevent the harm to patients that results from an incomplete understanding of the drug's mechanism of action.

This lack of mechanistic understanding lies at the core of the pharmaceutical industry's problems - a regulatory process that gets ever longer and more costly as it strives to fill the gaps as well as the devastation to both the industry's reputation and its bank balance that results from the nasty surprises that occur in clinical trials, or even after the drug has already been approved and sold to thousands or millions of patients.

For all of its shortfalls however, this empirical methodology has been a justifiable and rational approach for the pharmaceutical industry to take because simply put, biological systems are massively complex and there have until now, never really been any suitable tools with which biologists (or drug developers) could get a handle on this complexity. A civil engineer designing a suspension bridge is working with a system that is far less complex than even a single living cell and its behaviors and properties can be accurately understood by plugging the data into the models that physics and mathematics have provided the engineer. Biology alas, is replete with data, but woefully lacking in the kind of models that would allow its researchers to transform it into useful insight and to approach living systems (or the design of the drugs that modulate them) with an engineer's perspective.

Where computer science and biology intersect in the exciting field of digital biology, new approaches to computational modeling are emerging that have been specifically designed with the challenges of biological complexity in mind. These new methods allow biologists for the first time, to build real models of the massively complex systems inside living cells, without the constraints on scope and resolution that are inherent in the traditional approaches to biological modeling that have been borrowed and adapted from other fields such as physics and chemistry.

The ability to approach biological systems with a truly mechanistic perspective will transform basic research for academic biologists. For the pharmaceutical industry, it holds the promise of ushering in an era in which drugs are designed rather than discovered, finally lending an air of veracity to the industry's own TV commercials in which earnest researchers in white coats and safety glasses scribble designs for new molecules on perspex screens.

With the time and cost of drug development spiraling and the approval rate for new drugs actually in steady decline as R&D expenditure increases, something needs to change if we are not to return once more to measuring in human generations, the time it takes to discover a new drug.

The author Gordon Webster, has spent his career working at the intersection of biology and computation and specializes in computational approaches to life science research and development.

© The Digital Biologist | All Rights Reserved