I've been invited to present at the Extremely Large Databases (XLDB) 2012 conference as a practicing biologist who occasionally speaks with physicists, and I'm trying to come up with something to say that will explain why physicists and biologists don't often collaborate all that well.

Here are some guesses.

---

1. Biology has only weakly predictive theory.

Physicists tend to rely on a lot of math, because over the last two centuries it has turned out that relatively simple theoretical assumptions can lead to a surprisingly accurate version of reality. I don't think it would be wrong to say that you can view much of physics in the 1900s as a search for simple theoretical principles that correspond with observational data. The resulting theory has been an amazing success in many ways, and no one would even think of trying to interpret particle accelerator data without the aid of the Standard Model. In physics theory has turned out to be very predictive, in that you can make specific predictions based on numerical calculations and voila, upon looking you will find a match.

Biology, on the other hand, has no such strong theory. Evolutionary theory and ecological principles seem to describe some parts of biological reality quite well, but it is difficult to imagine coming out with very specific predictions from these theories and principles. Since evolution is the basis of biology (cue famous Dobzhansky, "Nothing in biology makes sense except in the light of evolution"), and evolutionary processes tend to construct stunningly functional solutions that nonetheless look like Rube Goldberg contraptions, a lot of biological detail looks like it has been created by an utterly insane genius.

There are lots of fun implications of this difference, including the fact that mathematical models are hard to apply in biology, and that a lot of biology inextricably involves what physicists (and the more frustratingly close-minded biologists) call "stamp collecting" -- looking for new data that doesn't fit whatever mental or mathematical or computational model you have. (RNAi, anyone?) I have news for you all -- stamp collecting is only boring when you have some freakin' idea of what the next stamp is going to look like!

Danny Hillis wrote an excellent article in 1993, "Why physicists like models, and why biologists should in which he relates a quote from Richard Feynman:

The late Richard Feynman, a physicist, once warned me never to compare physics to biology in front of a biologist: "It just makes them mad." He was joking, but he was also right.

The article goes on to talk about models, and models, and makes some really good points. Honestly I should just point people at that...

2. Biology is increasingly integrative, rather than reductive.

I've known some awesomely intelligent physicists, and even published with one or two. (I'm happy to name drop if anyone is interested...?) I've also known some really brilliant biologists. They are generally quite different types of people. Physicists tend to seek simple principles that they complexify as needed; biologists tend to seek broadly inclusive stories that they simply as necessary. If this is an accurate generalization, I betcha it comes back to the Rube Goldberg nature of biology: to extract general principles of evolution or ecology or development, you need to look at immense range of very confusing data; in physics, it seems to be possible to look at much smaller sets of data. Neither one is easy but I think the mental shift necessary to encompass biological behavior is tough for many physicists to make.

It is also the case that biology encompasses an immense range of disciplines, from ecology to molecular biology to animal husbandry. This is no different from physics, which runs the gamut from deep theory such as string theory, to solid state physics, to protein modeling, to... well, lots of things. I don't know why any physicist would expect a biologist to pick up physics in a few conversations, and I think the same expectation should be applied to physicists who are trying to pick up some biology: it's not going to be quick.

3. Career structure and lab management is just... different.

In physics, postdocs are considered larval faculty, and getting the PhD and moving on is much more of a selective step than in biology. In biology, getting a PhD can be easier -- often it involves a lot of mindless repetition in the lab, followed by writing the results up in collaboration with your advisor -- while the true selective transition is the postdoc to faculty step, which relatively few people make. (At this point I should perhaps mention that I have one of those biology PhDs, and yes, I did a lot of mindless repetition in the lab. I'm not knocking it; it's necessary. But it's different from what I would have done in most of the physics labs I considered joining.) Moreover, labs in biology tend to be run by one faculty member, whereas at least in my experience with physics at Caltech and Stony Brook and Ohio State, there is a tendency to work in larger groups with multiple profs.

These differences may stem from the labor-intensive nature of much of biology; the different funding approaches available and used in the different fields; the cultures of medicine and ecology and evolution vs the cultures of physics; etc. I'm not sure. But in my not-so-limited observation it's very different between biology and physics.

The most important effect, though, is that postdocs are not required to be very independent researchers in biology labs. They are beholden to their advisor, they aren't forced to do much "broad thinking", and they're often treated as semi-slave labor. Many of them push through this, either by force of personality or by luck of the draw in their boss, but many don't. This leads to science that is often driven by the "big boss", who is often a very important and well funded person, and necessarily limits the diversity of opinions and approaches.

I don't know how much of this contributes to the theorist for hire phenomenon, but I betcha it's a significant component.

4. Data generation and analysis is handled differently.

Narayan Desai pointed out to me a while ago that big data generation in biology is happening in individual labs, while often big data generation in physics is happening more centrally -- particle physics, astronomy, and the like, tend to have large instruments that generate the data, which is then filtered down to analysis groups. This is in part because of the robust collaborations between phenomenologists (theoreticians who analyze data) and experimentalists in physics, which relies on the strong theory that is available. In biology, however, individual labs and collaborations are generating big chunks of data, and they are generally being analyzed within those labs and collaborations. The lucky collaborations include good bioinformaticians and computational biologists, but there aren't enough to go around.

This leads to what I've been calling a "sick culture" of biology and bioinformatics, where there is a surprisingly small, low-diversity ecology of computational tools and approaches. You don't find anything like it in physics, as far as I can tell, and this is where I'd most like to adopt physics practices.

A friend who recently visited did point out that the sophisticated computational approaches as used in physics are no panacea -- he's used to attending talks about "robust approaches" that are obviously so tweaked that there's no way to tell if they're actually robustly useful.

---

That all having been said, I have been happily talking with and collaborating with physicists on my biology for ... decades, now. The most successful ones are the ones that approach biology with a certain amount of humbleness and willingness to learn completely new things; immersion therapy might be the best approach for those who have the time, as you can tell from looking at the careers of people like Hopfield and Delbruck. But that's not surprising -- the most successful biologists seem to be people who beg, borrow, and steal approaches from whatever field they need to.

I welcome input from people who have experience on one side of the fence or the other, or those who are straddling it. This is a tough subject to opine about and I'm happy to have my opinions be falsified or subjected to the light of alternate experience.

--titus

p.s. I'm writing this post from my own personal experience, and I'm certainly being chauvinistic to broad swaths of biology. Fields like genetics and ecology have been making use of models for decades. But the application of math and modeling to really squishy stuff like development and evo-devo has been very limited in utility, as far as I can tell. And we need it there, too.

p.p.s. Forgot to say -- thanks to Erich Schwarz and Ethan White for their comments on an early version of this post!