The Futile Cycle

I read a really interesting article today, penned by Scott Aaronson, a computer scientist at the Institute for Advanced Study in Princeton, New Jersey. He talks about NP-complete problems, methods that people have proposed for revolutionary advances in computing power, and how if human beings were able to create algorithms such that P=NP, then we would be like gods, in that we could replicate evolutionary history, predict the stock market, recreate Shakespeare, and so on. It’s an amusing article, requiring a bit of math, computer science, and physics to grasp, but entertaining nonetheless.

Someone asked me today whether, if I manage to get a graduate science fellowship from the Department of Defense, I would have any reservations about working “for the military.” If the military is interested in enough to pay me money, wouldn’t I be working on something that could potentially be used to harm people?

I don’t think so. Many parts of the government, from the NIH to the Department of Energy to the Department of Defense, sponsor basic science research, because they believe that promoting and culturing future scientists in research aligns with their overarching goals. The DOD states that it thinks that the promotion of science and engineering talent in the U.S. is essential for the country’s future security and world power.

In any case, any pertinence my research might have to defense or warfare would be indirect at best. It’s not like I’m working on developing neurotoxins or biological weapons. The closest I would ever get to working on actual products for the DoD would probably be biosensors that detect dangerous pathogens, or ways to disable them. I don’t really see that as an interesting research path, but maybe far into the future, I might get interested in it.

But in any case, most likely, if my research is interesting to the military, it would be because of some side effect of my research, or an unintended application. You can’t stop people from working on deadlier weapons, and people in the military will always be working on more effective ways of killing people. Basic science research has to go on and can’t be paralyzed by remote possibilities of unintended abuse. Almost anything that science produces can be perverted to use for destruction, from engineering virulent strains of bacteria and viruses to making nuclear bombs and finding key points in countries to maximize disruption. I can’t, say, stop researching ways to solve mathematical equations on computers just because the military might use those techniques to simulate explosions or bioweapons. Such knowledge could, say, lead to better ways to maximize food production, or lead to a better understanding of brain seizures, or make it easier to produce medicines. Equations don’t kill people; people kill people.

Science contains a sort of optimism that knowledge will improve society, and I believe it with all my heart. People now live longer lives, happier lives, and can do more with each hour of each day than those of times past. I will do the best I can to prevent people from killing people, but the inexorable march of knowledge continues to the betterment of mankind.

Lately, I’ve been thinking very much about science and applications of it, partly because I’ve been filling out fellowship applications that tend to ask me to justify my research interests, and partly from talking to Lester about various aspects of biomedical research, especially the NIH’s “roadmap”. The NIH funds a huge portion of the U.S. biomedical sciences research, almost $30 billion, which completely swamps the size of almost any other budget (NSF spends about $5 billion, Howard Hughes spends about $500 million, Gates Foundation spends about $800 million, and the Wellcome Trust spends $780 million). Because of this, the direction of basic bioscience research in the United States is strongly influenced by the NIH. For more on how the NIH funding process works, see the NextGen article (shameless plug).

One of the more controversial elements of the roadmap is the emphasis on translational research, or more colloquially, “bench-to-bedside” research, meaning basic science research that will quickly yield advances and applications to clinical medicine. Why controversial? Well, a lot of people, me included, feel that the private sector seems to be handling much of the necessary “translation” quite well. After all, although the NIH spends $30 billion on research, industry spends almost double that, and no doubt almost all of it is translational (spending on really fundamental science is essentially a “donation”, since they wouldn’t be expecting too much direct returns on that). Pfizer by itself spends more than $8 billion on research annually, and there are countless other biotech, pharma, biodevice, bioengineering, and other companies pouring money into research and development. All of it is private money. Each drug that comes to market generally costs about $1 billion in R&D spending. And these people are highly trained experts looking to move things into the clinic, not academics with some far-flung idea that probably won’t work. Thus, it seems like the NIH’s translational push is a solution in search of a problem. There’s nothing wrong directly with pushing for more applications for science, but there is only a limited amount of money to go around, and putting more money in translational research means that money gets taken away from the basic sciences, especially now that the NIH’s budget is getting cut.

What is basic science research? Well, for biology, it’s the real fundamental research, so the stuff on how evolution happens, how cells work with DNA, proteins, and RNA, how ecologies interact, how viruses work, and so on. Researchers look at how viruses enter cells, how bacteria sense chemicals, how plants grow, and what kinds of life exists at the bottoms of ocean floors. It’s a blurry line between biology basic science and “translational” research, because a lot of medical knowledge now comes from basic science research, from the microbe theory of disease, to the discovery of the structure of DNA (mostly done by physicists and chemists), to the development of NMR and MRI machines, and CAT scan technology, diagnostic mass spectrometry, antibody therapy, antiviral and antibacterial therapies, our understanding of neurological disorders, and so on. The line becomes especially blurry since basic science can also happen on human biology. Studying the pathways of human metabolism could be considered basic science, but it’s also essential knowledge for developing new treatments for diabetes, cardiovascular disorders, neurological disorders, hormonal disorders, nutritional disorders, aging, exhaustion, and so on. If all that’s necessary for medicine is a set of diagnosis trees to figure out “what’s wrong” and “what do I do to patch it up”, then perhaps this knowledge is unnecessary. But if one looks to the future, where new treatments must be made, then this sort of knowledge is essential, not only for those who work on the basic pathways in labs, but also for those who want to modulate the pathways for medical purposes.

Thus, basic science research today is the foundation for the medicine of tomorrow. It needs to be funded for our nations’ health, indeed, for world health. Sure, not everything in medicine is discovered by building directly on basic science, but mostly, those kinds of things are hard to promote. They’re shots in the dark, serendipitous because we don’t understand how they work, and so we don’t know how to build upon it for future medical research.

Let me speak of one example of basic science research. RNA interference (RNAi) was first discovered in 1990 with work on petunias (yes, the flowers). Later, in 1998, Mello and Fire discovered the mechanism of RNAi in a worm that eats bacteria, called C. elegans. This past year (2006), they received the Nobel Prize in Medicine for their work on RNAi. Why the prize in “medicine” for work that happened in flowers and worms? What does this have to do with humans? The medical community has panned RNAi for not having direct clinical applications; these critics are misguided.

For one thing, it was found that RNAi controls a lot of human embryonic development and is especially essential for neurological development. We can start to understand how single point mutations can change people in a very dramatic fashion. We used to think that it would alter the protein’s structure, as it did for things like sickle cell anemia, but we’ve now come to realize that most mutations often do something very different. In fact, most mutations, if they have any effect at all, can’t have an effect on protein structures, since only around 11% of the human genome codes for proteins. For example, in 2004 a child was born with abnormally large muscles (gross muscle hypertrophy, which means literally, “abnormally large muscles”), and they found that the single mutation caused the myostatin gene to be regulated by RNAi. The myostatin protein sequence itself was not changed; it was just silenced by RNAi. This kind of understanding is essential for expanding our eyes to the ways genes can be regulated, and thus how we can treat genes. Perhaps we can help muscular dystrophy by using RNAi to locally, temporarily suppressing myostatin.

For another thing, although direct clinical application of RNAi hasn’t worked so far, that doesn’t mean that RNAi doesn’t have clinical significance. RNAi has significantly sped up biological research. Now, for example, instead of having to create mutated mammalian cells to investigate the functions of certain genes, one can use RNAi knockdown in order to quickly generate the same effect without the trial and error of actual mutations, which can kill the cells (rendering them useless or at least really difficult to work with) and are difficult to generate. RNAi allows for larger scale genomic research, to scan many hundreds of genes simultaneously in order to screen for what causes certain diseases or ailments.

Let me give another example. Gene therapy was deemed too risky and not effective a long time ago, and it was pretty much abandoned for the time being, at least until technology gets better. But does that mean that modulating genes (transgenics) is useless? Of course not! This technique is used all the time in laboratories order to study biological diseases. To say it has no clinical relevance is just plainly ignorant; indirectly, it has contributed enormously to medicine.

Here’s yet another example. Green Fluorescent Protein was isolated from jellyfish, and it has the remarkable property that, when exposed to a bit of oxygen in solution, it’ll start to fluoresce green. Great, so now what, make green glowing pigs? What’s the point? Well, GFP is now an essential protein used to study the expression of many genes, tracking their locations in cells, and even using them to discover how neurological genes work. Sure, making things glow green will probably never have a direct clinical purpose, but it’s a huge advance in medical research.

I think part of the problem with some people is that they expect technology to revolutionize things all the time, or work instantaneously. What these people don’t realize is that most new technology is actually made to enhance our use of existing technology. I mean, we still take pills to cure many of our diseases. Ultimately, in the future, a lot of our medical therapies will still be based on taking pills. They’re (relatively) cheap and easy to take. What will be different in the future is how we make those pills and discover the stuff that goes in them. Before the advent of nuclear research and the development of radioactive isotopes, we wouldn’t have been able to develop drugs like Gleevec, for which we need to use radioactive isotopes in order to verify their effectiveness in the laboratory. Cheap ways of synthesizing DNA in the lab has allowed for technology like PCR and the creation of DNA/RNA probes for biological tests of drugs. New ways of synthesizing chemicals allow us to make drugs that might be infeasible to make without them. RNAi will allow us to choose which enzymes to target with drugs by quickly analyzing the effects of knocking them out. Understanding human metabolism helps us pick good targets for drugs. Research on fruit fly embryonic development proteins can lead to key insights on inflammation in humans, which gives us targets for drugs.

In addition, basic science takes some time to translate into biomedical applications. And we have had a very, very bad record for predicting what research will lead to good clinical results. Thus, I believe in the “cast the net widely” philosophy for basic biomedical research. When good ideas come, we run with it, but you can’t force ingenuity. All you can do is cultivate it by giving it the soil it needs to grow. Without the soil, the plant will grow taller for a bit, looking for light and nutrition, and then die. Basic science needs to be funded. Translational research is already having tons of money being poured into it. I say that we should leave translational research to the experts and the investment of those looking for profits, and have the government pour money where it can have the greatest impact, which is basic science.

A recent dorky poll on favorite science textbooks has been launched on Uncertain Principles. They had a great quote in the comments of that post:

“The problem with [The Feynman Lectures on Physics] is you read it, and you say ‘Yes! I understand this! I’m doing physics!’ And then you try to solve a problem, and you find that, well, that you’re not Feynman.”

Anyway, I thought I’d enumerate my favorites here a bit more extensively than comment space would allow.

Evolutionary Dynamics, by Martin Nowak: Great explanation of evolutionary biology. After reading this, you realize that without mathematics, any discussion of evolution becomes a series of just-so stories and hand-wavy explanations. With math, however, you can do wonderful things like look at the time scale of evolution, the effects of kin selection and group selection, effects of population size, which societal structures that allow for cooperation, the necessity of having universal grammars for human language, the development of coherent languages, the evolutionary effects of newly introduced language structures, the natural development of cancer, the evolution of parasites and destructive pathogens, the etiology of HIV, and so on. This kind of course seems necessary for a well-rounded scientist, and I don’t know why, 80 years after the modern evolutionary synthesis promoted by Fisher, Wright, and Haldane, that this kind of mathematical teaching isn’t a requirement for all biologists. Sure, it’s not necessarily pertinent to the day-to-day research that goes on in many biology labs, but it’s necessary for the same reason that basic physics needs to be taught for chemistry and biology research. It’s about the fundamental laws of nature, and any scientist without at least some exposure to this kind of course is sorely lacking.

Thermal Physics, by Kittel and Kroemer: A great introductory statistical thermodynamics book. The derivations are simple, the applications and examples are numerous, and the exercises are challenging, but interesting. There’s a lot to be praised about their early introduction of simple ideas with powerful results, such as the partition function. Still, a caveat: at times, the content can be a bit confusingly organized, in that certain fundamental topics are scattered in many different places, de-emphasizing their importance. This may be more salient to me because I hadn’t taken a formal classical thermodynamics course beforehand, so I didn’t know about such important things as the Maxwell relations, which are only briefly mentioned in passing on one page of this book.

ANSI Common Lisp, by Paul Graham: I haven’t finished it yet (not even close), but so far, it’s a very readable introduction to Lisp, which is a pretty neat programming language. It’s normally pretty hard to learn, because you have to think differently when you want to program in it (as opposed to, say, C or C++), and it’s a bit difficult to pick up at first. This book makes it a lot less painful, and gives good exercises for the reader.

Advanced Mathematical Methods for Scientists and Engineers: Asymptotic Methods and Perturbation Theory, Bender and Orszag: This is a wonderful mathematics textbook on how to do approximate analysis of dynamical systems. A good mix of hard and easy topics. Very useful, too.

A Genetic Switch by Mark Ptashne: This textbook is a very nice overview of one marvelous biological system, the lysis repressor switch of bacteriophage lambda. It’s a great, very short book with lots of pictures and diagrams, and very understandable. A bacteriophage is simply a virus that attacks bacteria, and many of them often have a system in which the virus is latent in the bacterium until some stimulus (often UV damage of the bacterial DNA) triggers it to start reproducing “lytically” (i.e. until the bacterium bursts). Biology had to solve a difficult engineering problem, which was to make the switch quick and sensitive to the stimuli you want, but robust enough that it wouldn’t accidentally be triggered. This book details in a very easy and colloquial fashion how evolution has solved this problem in the case of phage lambda. It’s fascinating to see how each additional layer of complexity on the genetic switch serves to fix a slight problem, to tweak the system and make it just a bit more efficient. The only problem is that at times, the book assumes a bit of molecular biology knowledge that the intended audience (intro high school and undergraduates) may not have.