So Sir Paul Nurse gave a talk today where he discussed the 5 big ideas in the Life Sciences.
Instead of going through the talk, I'll just say that he ended it with his 5th big idea - biological organization. It reminds me of discussions I've had with colleagues on cell polarity. And in someways it's no surprise as this is Paul Nurse's current topic of interest. To generate a front and a back, cells must have a large interactive cellular network full of scaffolds and feedback loops that enforce a differentiation within a cell. Cell polarity cannot be explained simply in describing the polarity of molecules, you can only get insight by looking at how the system works in toto.
Coming back to Paul Nurse, after his talk he fielded questions, and so I asked:
The chemistry within cells relies on two types of networks, one set is robust and self organizing like the cytoskeleton or the cell cycle, the other is a loose connection of factors, each of which can have profound effects on how the first network operates. How fruitful can modeling of these networks be if we do not yet have all the components?
Nurse's answer was that we could find all sorts of reasons not to perform experiments, but that we should try anyway. In a way I agree with his answer, but I guess I worry about whether our models have ALL of the necessary components. Don't forget that given enough variables we can construct any model to fit the data. I am constantly amazed at how little we actually know about the inner workings of a cell. And I am amazed how so many people in the biological sciences don't realize this. So in retrospect I should have asked:
Given that we do not yet know what all the molecular players are within cellular networks, how do we know that any model is a valid representation of what is happening in cells, or whether our models give the right answers for all the wrong reasons?
Ironically someone then stated that the current paradigm of cellular networks depicts molecules within a cell as activating each other in a linear succession and that complicated messy feedback loops are hard to study and even harder to publish within the current framework of scientific publishing. Paul Nurse thought that there were enough individuals interested in systems biology that the science would get out and the truth will be uncovered eventually, but that it might take a long time and that his career might be destroyed in the process. In other words, we don't know that this approach will work, go try experiments, but be prepared to fail. Honestly if was a senior PI with nothing much to loose, I might try my hand at Systems biology, but for a young scientist why go there when there are lots of fundamental properties within cells yet to be discovered?
- Log in to post comments
The first reason that comes to my mind is that physicists have had lots of luck studying things without knowing all the players, coming up with situations that imply another player might be involved, and then successfully finding that other player based on predictions made by theory.
The "big picture" of the cell is far more interesting to some people than the individual players, just as an orchestra can be more interesting to listen to than a single musician with just one instrument. That isn't to say single instruments can't make beautiful music, but it seems obvious to me why studying the entirety of the cell (with imperfect knowledge) could certainly be more interesting to some than a well-characterised subsystem within the cell.
What with systems biology being all the rage, I have a hard time believing that it's difficult to publish papers about feedback loops. For example, look at all the papers published by A. van Oudenaarden, Uri Alon, Adam Arkin etc.
[Maybe I'm misunderstanding your comment]
PhilipJ,
Evolution is a diversity generator, thus biology is full of weird complicated machines that have impacts on the self organizing machines that form the basis of cellular life. For example there are a plethora of proteins that bind and modulate the activity on the tips of growing microtubules. You can try to model the behavior of a microtubule array, but without knowing what all the tip modulators are and what they do, your model is useless. These modulators are the reason a cell can look like a sperm, an inner ear cell, a retinal cell, an oligodendrocyte or a hepatocyte. To really understand how those cells work it is crucial that we nail down all the important players that modulate the cell's core processes. Also I want to say that the goal of science is not beauty (although much of what we study is beautiful in my opinion) but insight - or a deeper understanding of what is going on.
Alex,
I agree! That last question posed to Nurse was by someone else. Systems biology is hot and if you have a half decent paper I'm sure that you would have no problem publishing it. Some I have to admit have made some headway - I'm thinking of Peter Sorger's work. However is it right? Does it give us any valuable insight? Sorger, when analyzing apoptosis, did not factor in the turnover rates of his proteins. So sure his model worked, but is it close to what is really going on? His disregard of protein turnover is a dangerous assumption to make - and it gets too complicated to factor it in the model. And if a new major component of apoptosis is discovered, then what?
So in a way Paul Nurse is right on two accounts, we shouldn't stop trying, but it'll probably be a while before we get insight from this aproach.
I saw you!
Re: the usefulness of systemy things: I really thought math modeling was total bullshit until I took a bioengineering class at MIT. I realized that while a lot of it is bullshit - I mean, you just make up numbers if you don't know real kinetics, and the modeling people don't understand that sometimes you just can't really know - it can actually be really useful for the formulation of novel hypotheses.
For example, you can make up a model for a system that has a lot of "moving parts", such as multiple GTPases, based on what you do know and run simulations where you vary different parameters. If you see anything that resembles real life, it gives you a clue of where in the system to look for missing key players. If you don't see anything that resembles real life, then you can play around with what you would need to input into the system to get it to look like real life and in that way can ask new and interesting questions about an old system.
I did think his points about considering network dynamics, spatial, and temporal levels of regulation were spot-on. But I am biased because that's kind of the meat of my dissertation . . .
I talked to him for a while after at the reception and he was super cool. Did you know he proof-read The God Delusion for Dawkins?
Well since I am now recognizable, you'll have to introduce yourself the next time we bump into each other. (I'm guessing that you are in S.T,'s lab - we once got paxilin -/- cells from your boss when i was in the Gundersen lab, I've never met her though ... )
Back to Nurse ... I guess I agree that to really understand how cellular functions work we'll have to move to modelling. But to do so we need to know quite a bit about what is going on. My question is more along the lines, are we there yet? I guess (as Paul Nurse said about Mendel) we can attempt to find something that works, and maybe that's good enough for now ...
Oh! You were in the Gundersen lab? Cool. We should definitely introduce ourselves in real life.
ST is my collaborator and on my committee but I'm not in her lab. (Maintaining some guise of anonymity.)
I have to say, though, today I am hating all things having to do with paxillin and its associates. Grr. Grrrr. Grrrrrr.
Never spend Saturday in lab, it will all just go terribly wrong.
"Ironically someone then stated that the current paradigm of cellular networks depicts molecules within a cell as activating each other in a linear succession and that complicated messy feedback loops are hard to study and even harder to publish within the current framework of scientific publishing."
People who study cellular feedback loops that participate in circadian timekeeping do not have trouble publishing their work. In fact, that is the dominant paradigm in the field--unduly so, in my opinion.
--ttcc
"Don't forget that given enough variables we can construct any model to fit the data."
Yep, agreed.