An open-source shot in the arm?
June 12, 2004
Medicine: The open-source model is a good way to produce software, as
the example of Linux shows. Could the same collaborative approach now
revitalise medical research too?
CAN goodwill, aggregated over the internet, produce good medicine?
The current approach to drug discovery works up to a point, but it is
far from perfect. It is costly to develop medicines and get regulatory
approval. The patent system can foreclose new uses or enhancements by
outside researchers. And there has to be a consumer willing (or able) to
pay for the resulting drugs, in order to justify the cost of drug
development. Pharmaceutical companies have little incentive to develop
treatments for diseases that particularly afflict the poor, for example,
since the people who need such treatments most may not be able to afford
It is in this environment that a number of medical biologists,
lawyers, entrepreneurs and health-care activists have sought
improvements. They have suggested borrowing the "open-source" approach
that has proven so successful in another area of technology, namely
software development. This is a decentralised form of production in
which the underlying programming instructions, or "source code", for a
given piece of software are made freely available. Anyone can look at
it, modify it, or improve it, provided they agree to share their
modifications under the same terms. Volunteers collaborating in this way
over the internet have produced some impressive software: the best-known
example is the Linux operating system. So why not apply the open-source
model to drug development too?
In fact, open-source approaches have emerged in biotechnology
already. The international effort to sequence the human genome, for
instance, resembled an open-source initiative. It placed all the
resulting data into the public domain rather than allow any participant
to patent any of the results. Open source is also flourishing in
bioinformatics, the field in which biology meets information technology.
This involves performing biological research using supercomputers rather
than test-tubes. Within the bioinformatics community, software code and
databases are often swapped on "you share, I share" terms, for the
greater good of all. Evidently the open-source approach works in
biological-research tools and pre-competitive platform technologies. The
question now is whether it will work further downstream, closer to the
patient, where the development costs are greater and the potential
benefits more direct.
Open-source research could indeed, it seems, open up two areas in
particular. The first is that of non-patentable compounds and drugs
whose patents have expired. These receive very little attention from
researchers, because there would be no way to protect (and so profit
from) any discovery that was made about their effectiveness. To give an
oft-quoted example, if aspirin cured cancer, no company would bother to
do the trials to prove it, or go through the rigmarole of regulatory
approval, since it could not patent the discovery. (In fact, it might be
possible to apply for a process patent that covers a new method of
treatment, but the broader point still stands.) Lots of potentially
useful drugs could be sitting under researchers' noses.
The second area where open source might be able to help would be in
developing treatments for diseases that afflict small numbers of people,
such as Parkinson's disease, or are found mainly in poor countries, such
as malaria. In such cases, there simply is not a large enough market of
paying customers to justify the enormous expense of developing a new
drug. America's Orphan Drug Act, which provides financial incentives to
develop drugs for small numbers of patients, is one approach. But there
is still plenty of room for improvement-which is where the open-source
approach might have a valuable role to play.
In a paper presented this week in San Francisco at BIO 2004, the
Biotechnology Industry Organisation's annual conference, Stephen Maurer,
Arti Rai and Andrej Sali -two lawyers and a computational biologist,
respectively-called for an open-source approach to invent drugs to fight
tropical diseases. It would work like this: a website they call the
Tropical Disease Initiative would allow biologists and chemists to
volunteer their expertise on certain areas of a specific disease. They
would examine and annotate shared databases, and perform experiments.
The results would be fully transparent and discussed in chat rooms. The
authors expect that the research, at least initially, would be mainly
computational, not carried out in "wet" laboratories.
The difference between this proposal and earlier open-source
approaches in biomedical research is that where before scientists
swapped software, here they would collaborate on the data. And where
projects such as the mapping of the human genome relied on massive
top-down government involvement, this proposal would, like an
open-source software project, be the result of bottom-up
self-organisation among researchers themselves. That said, the authors
acknowledge that a government or grant-giving charity would probably
have to provide the initial funds.
Moreover, the results of the research would not be made available
under an open-source licence of the kind that governs software projects.
Instead, the final development of drug candidates would be awarded to a
laboratory based on competitive bids. The drug itself would go in the
public domain, for generic manufacturers to produce. This, the authors
state, would achieve the goal of getting new medicines to those who need
them, at the lowest possible price. "We are so used to patents that we
forgot ways to discover drugs in the public domain, and we need to
rediscover them," says Mr Maurer, of the Goldman School of Public Policy
at the University of California in Berkeley.
This is just one of many attempts to extend elements of the
open-source software-development model to drug research. Yochai Benkler,
a law professor at Yale, imagines test-tube and animal studies organised
in this manner, exploiting the "excess capacity" of graduate students
and university labs, much as students and academics also contribute to
open-source software development.
Eric von Hippel, a professor at the Massachusetts Institute of
Technology's Sloan School of Management, is investigating how secondary
uses for drugs are discovered, with a view to harnessing doctors and
patients to record data. Many medications are approved for one purpose,
but are regularly prescribed for another, "off-label" use. In many
instances, new uses for a drug are discovered only after it is on the
market, when a sort of natural experimentation takes place. For
instance, Botox was approved in America for treating eye-muscle
disorders, and only later found to remove wrinkles. In Europe and
America, as many as half of all drug prescriptions for certain diseases
fall into this category. The drugs often do not go through the formal
process for other uses because the cost of regulatory approval is so
This is a problem for a number of reasons. First, it means that drug
companies are prohibited from advertising the medications based on these
additional uses, so some patients may not get the treatment that would
benefit them. Next, insurance companies in America usually only cover
on-label use. And the effectiveness of the treatment is not formally
evaluated. Dr von Hippel's idea is to decentralise the process of
obtaining data on the off-label use, by collaborating with volunteer
doctors and patients. By defraying costs in this way, it might then be
possible to obtain regulatory approval. It is, in effect, an open-source
clinical trial. Because the drug has already been approved, it has
passed first-phase tests for safety. These do not have to be repeated.
Second and third-phase drug-approvals test for efficacy and
side-effects-and these are the very areas where getting formal approval
for off-label use is sensible.
Meanwhile, not far from Dr von Hippel at MIT, thousands of fruit
flies are being decapitated. Peter Lansbury, the head of a research lab
at Harvard Medical School, avows that they are treated with chloroform,
so "they don't feel a thing ". The fruit flies have Parkinson's disease,
and Dr Lansbury's research is examining the therapeutic effect of a
thousand approved drugs, on which the patent has expired in most cases.
Might one of them turn out to be an effective treatment?
This sort of research is unusual because there is no working
hypothesis to prove and no way to profit if the project is successful.
It has simply never been studied before, and should be, says Dr
Lansbury, who is the co-founder of the Laboratory for Drug Discovery in
Neurodegeneration. The laboratory has around 25 researchers and an
annual budget of $2.5m to work on neurodegenerative diseases, such as
Parkinson's or Huntington's, to which the major commercial drug
companies devote few resources because their potential market is small.
Dr Lansbury refers to the work as "not-for-profit drug discovery",
but he sees direct parallels with the open-source approach. For one
thing, his group places much of its data in the public domain. Secondly,
though the research is mainly happening among different research labs
within the confines of Harvard at the moment, the goal is to involve
other scientists around the world. Only through this sort of
collaborative, distributed approach will treatments be found for these
diseases, he says. As for the intellectual property that may be created,
the goal is to use patents only to license treatments cheaply to
pharmaceutical companies to ensure a supply of drugs at low cost. But
the most important thing is to discover the drugs in the first
place-something commercial drug-development seems unable to do.
There are a number of other similarities between biomedical research
and open-source software development. First, both fields attract the
same sort of people. Biology, like software, relies on teams of
volunteers, notably graduate students and young professionals, who have
an incentive to get involved because it will enhance their professional
reputations or establish expertise. Both medical biologists and computer
scientists aim to improve people's lives and make the world a better
place. And as the human-genome project showed, both cultures respond
strongly to grand projects, not just financial incentives-possibly
because they are generally highly paid to begin with.
That said, the dissimilarities are profound. The financial needs and
time to complete projects are wildly different. A new piece of software
can be thrown together in days or weeks, and rarely more than a few
months. The barriers to entry are low: many pieces of software begin
life in an enthusiast's bedroom or garage. Pharmaceutical research, in
contrast, is measured in years, fails more often than it succeeds, and
requires hard-core credentials and in many cases expensive equipment,
not just hard work.
Moreover, the computational portion of the drug-discovery
process-typified as upstream, far from the patient, at the early-stage
level, where profits are thinner-is not the costly bit. Rather, it is
the less computer-intensive things such as toiling in wet laboratories,
performing clinical trials and navigating the regulatory-approval
process where one finds the bulk of the cost of bringing a drug to
market. The closer to the patient one goes, the tougher it is to imagine
open-source processes making a significant impact.
The application of the open-source approach to drug development may
prove to be more useful as an analogy than an application, notes Janet
Hope, a lawyer completing a doctorate on "open-source biotechnology" at
the Australian National University, in Canberra. One reason is that
different intellectual property rights apply, and are protected
differently. Software usually falls under copyright, which arises
automatically and without cost to the author. Biomedical discoveries are
generally protected by an entirely different legal regime, patents,
which are costly to obtain.
This helps explain why the drug-discovery and development projects
place their work in the public domain, rather than trying to enforce
some form of reciprocal openness through an open-source licensing
agreement, as software does. Those involved in the human-genome project
investigated the possibility in 2000 of applying an open-source
licensing agreement to the results, but decided that simply throwing the
results into the public domain-without any restriction on their use-was
better. Its successor project, the International HapMap Project, which
is mapping the common patterns of variation within the genome, imposes
an open-source licence for research in progress. But it places the
completed data in the public domain and allows patents on subsequent
This suggests that continued reciprocal sharing, a key part of
open-source software development, may not have a meaningful equivalent
on the biological side of the fence. With open-source drug discovery in
the public domain, where there is no legal obligation to share one's
inventions, there is no guarantee that philanthropic sentiments will
override self-interest. Participants can always choose to send their
results to the patent office rather than the communal web site. While
the open-source approach shows much promise in drug discovery, it is
certainly no panacea.
More broadly, two big questions remain unanswered as the open-source
approach starts to colonise disciplines beyond its home ground of
software development. The first is whether open-source methods can
genuinely foster innovation. In software, all that has been developed
are functional equivalents of proprietary software-operating systems,
databases, and so on-that are sometimes slightly better and sometimes
glaringly worse than their proprietary counterparts. Their main
distinction, from users' point of view, is simply that they are
available free of charge. Curiously, this matches the complaint levelled
against pharmaceutical companies for developing "me-too" drugs to
compete with other firms' most successful product lines-witness the
current crop of Viagra imitators-rather than spending their research
money in an entirely new area.
The second question is semantic. What does it mean to apply the term
"open source" in fields outside software development, which do not use
"source code" as a term of art? Depending on the field in question, the
analogy with source code may not always be appropriate. It seems the
time has come to devise a new, broader term than "open source", to refer
to distributed, internet-based collaboration. Mr Benkler calls it
non-proprietary peer-production of information-embedding goods. Surely
someone, somewhere can propose something snappier.