Developing Unifying Theories for Biology

Top

00:00:18.09 and I want to talk to you today about some of the challenges
00:00:22.02 that we face at the interface between physics and biology
00:00:26.02 I think many of you know that there's a lot of excitement at this interface because
00:00:30.25 more and more of biology today is becoming accessible to the quantitative experiments
00:00:36.19 so there's a classical descriptive biological tradition
00:00:40.22 but more and more of the phenomena that we see in the living world
00:00:44.27 are now susceptible to the kinds of quantitative large-scale experiments
00:00:49.10 that are the norm in the physical sciences
00:00:52.14 and this transition has created a lot of excitement
00:00:55.15 it produced an enormous amount of data
00:00:57.18 and in areas of biology that range from the behavior of a single molecule of all the way up
00:01:02.22 to the dynamics of large populations of organisms
00:01:05.23 we' ve seen sort of independent evolutions in every sub-field
00:01:09.28 where the whole character of how we approach the phenomena of life is changing
00:01:13.24 And what I want to talk to you about today is this
00:01:17.20 transition from a classical descriptive science to
00:01:23.25 a quantitative science being able to summarize what we observe
00:01:30.19 in mathematical models and finally I want to talk about the challenge of reaching one step further
00:01:36.25 to doing the kind of thing that we're used to and theoretical physics
00:01:39.25 but has not really happened yet and biology
00:01:42.25 and that's to have more than just the models for individual phenomena
00:01:48.04 often with lots of parameters that you have to fit in order to get all the details of the day to right
00:01:53.16 but rather having theories that are a bit more grand in scope
00:01:57.24 and have the power to predict
00:02:00.20 what you will see iwhen you go out and look at the world
00:02:04.17 from some more basic principles
00:02:07.01 and it's this search for a kind of unifying mathematical language
00:02:12.05 that I think has been characteristic of physics
00:02:14.22 and which if you said that you were interested in doing this in a biological context
00:02:20.00 a generation ago for example when I was a beginning graduate student
00:02:24.23 people looked at you with a lot of skepticism to put it politely
00:02:29.26 that you noticed it just didn't seem that that biology was susceptible
00:02:33.11 to that kind of theorizing and this revolution on the experimental side I think has changed everything
00:02:40.02 and has made the search for that kind of theory theory in the sense that we understand it
00:02:44.22 in theoretical physicis a real research program rather than just a fantasy
00:02:48.21 about how we might someday be able to understand things
00:02:51.18 tso the way I'm going to try and introduce you to this to this evolution of our understanding
00:02:56.18 and then the challenge that lies in front of us
00:02:59.20 is to take a very specific example which is the control of gene expression
00:03:03.29 and in particular the transcriptional control of gene expression
00:03:06.16 and with something that's happening in every cell in the living World all the time
00:03:11.16 and I know digging in a little bit into that example
00:03:14.25 and show you where I think the challenge lies
00:03:18.06 and at the end on soon out
00:03:19.23 and know that that many of the things that were saying about that particular example
00:03:24.02 really were things that you could be saying about a much wider range of biological problem
00:03:28.02 so you'll see that this challenge of building theory
00:03:31.00 is something much more general
00:03:33.23 So the example is transcriptional regulation
00:03:36.14 you remember that every cell in your body has the same DNA
00:03:40.21 what makes the self difference is that they will read out different genes
00:03:44.06 to make protein this is happening not just in us but in
00:03:48.03 single-celled organisms as well when bacteria find themselves in different environments
00:03:52.26 they read out the information coded in different genes
00:03:56.10 to make different proteins that are useful in those different environments
00:03:59.04 and that process is controlled at many levels
00:04:02.15 but one of the control steps
00:04:04.13 is right at the point where you read out the information
00:04:07.21 that's coded the DNA synthesized messenger RNA
00:04:10.09 So what you should keep in mind is a stretch
00:04:15.12 of the genome, a piece of DNA
00:04:18.13 and along this there's the gene
00:04:23.24 that is coded in the coding region
00:04:26.12 for the particular protein that you're interested in
00:04:28.15 you know what's going to happen when this information is read out
00:04:32.22 during transcription there's a large transcriptional apparatus
00:04:37.09 that sits down on the genome and starts to walk along the DNA
00:04:43.18 as it does so, it spits out a molecule of messenger RNA
00:04:50.29 that finds its way out. And eventually of course
00:04:56.14 is transported to the ribosome where it's translated in protein
00:05:01.09 you can regulate the amount of protein that you make as many stages
00:05:08.00 One of them is just to regulate this process
00:05:10.14 So there're also binding sites on the genome
00:05:14.15 that sits here close to the point where transcription is initiated
00:05:22.07 where other proteins which are called transcription factors
00:05:26.27 or just called TF can bind
00:05:31.02 of course there are many of these proteins also floating around in solution
00:05:35.03 and when these proteins bind, you can imagine that if they find and get in the way
00:05:44.10 of the transcriptional apparatus then they will repress the expression of the gene
00:05:49.14 alternatively they might bind in such a way as to make it easier
00:05:53.01 for the transcriptional complex to find its way to this start site
00:05:57.03 and get going in the process of transcription
00:06:00.05 So you can have transcriptional repressors, you could have transcriptional activators
00:06:03.29 Now the simple geometrical picture is not bad if we're talking about bacteria that's where it came from
00:06:11.04 by the time you get to complicated organisms like us
00:06:13.25 this is probably a over simplified cartoons
00:06:16.23 but it's nonetheless of the thing that people keep in mind
00:06:19.03 and the results of this process is that if you think about how much
00:06:25.29 of the gene you're actually expressing and you plot that
00:06:34.06 as a function of the concentration of the transcription factor proteins in solution
00:06:44.14 then if this is a repressor so it gets in the way of transcription
00:06:49.23 you expect that as you have more and more of the transcription factor
00:06:52.16 you fill up more of the site and blocked transcription more effectively
00:06:56.24 and what you'll see is that transcription falls
00:07:00.29 as the function of the transcription factor concentration
00:07:06.19 Alternatively if this is a transcriptional activator
00:07:11.19 then you expect that as more and more of the transcription factor
00:07:14.18 is present in solution more of it binds and it makes it easier and easier for the transcription process to get started
00:07:19.29 and so you see something that looks like this
00:07:23.21 Now in the old days this is where you have to stop
00:07:30.18 but now you can do much better. You can reach into the cell
00:07:36.03 You can engineer the genome so that transcription factors have attached to them
00:07:43.01 a fluorescent protein and so you can look at the cell under a microscope
00:07:47.22 and measure of the amount of fluorescence and that will tell you what the concentration of a transcription factor is
00:07:52.26 similar way you can engineer the gene that's being read out
00:07:57.05 also to have a fluorescent protein attached to it perhaps in a different color
00:08:01.05 and then you can measure not only the concentrations of transcription factor but also
00:08:05.23 the expression level just by measuring the forescence
00:08:08.29 now convincing yourself that that's a reliable measurement and really proportional
00:08:12.20 to how many molecules you have that takes work but that's been done in many cases
00:08:17.10 and so these relationships, these sort of input-output relationships
00:08:20.19 between the control signal and the output of a transcriptional process of
00:08:25.01 the whole process of gene expression has now been measured quantitatively in many examples
00:08:30.22 and one of the things you learn when you do that
00:08:32.25 is that even if you find two cells that have the same concentration of the transcription factor
00:08:38.28 or even the same cell at a different moment in time
00:08:41.23 comes back to have the same concentration of the transcription factor
00:08:45.09 what you see is that the actual level of gene expression fluctuates
00:08:48.26 and that fluctuation have you do enough work in the experiment
00:08:52.12 you could convince self is not coming from noise in your measurement
00:08:55.20 it really is variability and the behavior of a cell itself
00:08:58.22 and so these pictures should be associated will error bars along the curve
00:09:06.08 and those error bars really represent noise in the process
00:09:12.23 of gene expression and its control
00:09:15.04 And we understand a very basic physical reason why this happens
00:09:19.06 Individual molecules behave randomly
00:09:21.17 they take random walks through solution and find their way to their binding sites
00:09:25.24 There aren't very many of these molecules altogether
00:09:28.24 if you look inside a bacterial cell or the nucleus of a eukaryotic cell
00:09:34.03 you'll find that the total number of transcription factor molecules might be a order of 100
00:09:38.19 and it's not only that the cell can't decide
00:09:42.00 to account all of the molecules inside it
00:09:44.03 those molecules have an impact only when they find their way to this binding site right here
00:09:49.13 in this very specific points along the genome
00:09:52.01 which of course is may be altogether 100 base pairs
00:09:56.11 out of millions in a bacterium and Orders of magnitude more in eukaryotic cells in us
00:10:02.19 And so this very small number of molecules behaving randomly
00:10:06.16 the randomness does an average out
00:10:08.13 the way you were taught in your high school chemistry course
00:10:12.08 In high school chemistry we talked about the concentrations of molecules
00:10:15.18 as if that was a pure number
00:10:17.29 but really there is a limited number of molecules in any small volume
00:10:21.24 when you do experiments in a beaker that has a liter of material
00:10:26.20 you don't notice this. But when you look inside a bacterial cell
00:10:30.28 which is 10 to the -12th, a thousandth of a billionth of a liter
00:10:34.28 then these effects to come really really important and that's where these error bars come from
00:10:39.05 So we've been able to push to the point where we have
00:10:44.16 mathematical models that describes this process
00:10:47.09 Essentially you take this picture seriously molecules bind and unwind from their binding sites
00:10:52.14 When they bind, they influence the binding of the transcriptional complex
00:10:55.28 the transcriptional complex moves at some rate
00:10:59.08 it synthesizes messenger RNA molecules
00:11:01.18 they're degraded at some rate and so on
00:11:03.21 and when you're done you can actually produce models that will describe
00:11:08.04 this behavior and really there's three parameters although there's lots of things that go into determining this
00:11:13.16 There's three things. There is the overall level of expression
00:11:17.00 which has to do with how fast you can make the messenger RNA molecule
00:11:20.08 how long they live, how efficiently are translated into protein
00:11:23.12 There is the scale on the x-axis here
00:11:26.06 How much transcription factor do you need in order to turn the process on halfway
00:11:31.03 Ultimately that's controlled by the binding, the strength of binding of the transcription factor molecules
00:11:37.06 to these particular sites along the DNA
00:11:39.11 And then is the steepness of the curve
00:11:41.27 and that really reflects how many of these molecules cooperate
00:11:45.04 If one transcription factor molecule binding
00:11:48.09 drives the whole process, then this curve will be very shallow
00:11:51.17 But if it were to turn on transcription, you need three molecules to bind at once
00:11:56.16 or 5 or 7 which is what you see in real examples
00:11:59.20 then this process can become very steep
00:12:01.24 This curve can become very steep
00:12:03.09 and if it is almost an on-off switch
00:12:05.19 Now we went from description the measurement to
00:12:11.05 mathematical models that can describe this curve
00:12:13.01 we can even understand where these error bars come from
00:12:15.12 So what more do we want?Well, the problem is that these models have parameters in them
00:12:21.07 The halfway point along as curve is determined by how
00:12:27.14 tightly the transcription factor binds to this site
00:12:30.14 but how tightly the transcription factors bind to this site
00:12:32.21 depends on the structure of the transcription factor
00:12:35.14 but also on the sequence of the DNA in this particular region
00:12:38.08 The same transcription factor could also be controlling a different gene
00:12:42.00 and you would discover that this halfway point is in a different place
00:12:45.14 because the sequence of long here is different
00:12:47.24 so that means that these parameters are not fixed forever
00:12:51.01 In fact they evolved. And they evolve rapidly
00:12:54.18 because they're in the regulatory regions of the genes
00:12:56.29 And in fact it's even worse than that. The transcription factor proteins themselves can be modified
00:13:03.15 they can be phosphorylated, they can modified in other ways
00:13:05.26 which also affects the strength of their binding
00:13:08.04 So these parameters are things that have been chosen
00:13:11.05 by evolution. They've been chosen by the physiology of cell
00:13:15.03 and so if you have a model that ends by having these parameters
00:13:19.16 you've left open a big question
00:13:21.00 Why are these parameters what they are
00:13:23.08 What is it about this particular choice but advantageous for the organism
00:13:27.28 and that's a problem not only in describing one of these prophecies
00:13:32.13 but normally these processes are embedded in a network
00:13:35.21 There are many genes, a single transcription factors could affect many genes
00:13:42.08 but those genes themselves could code for transcription factors
00:13:46.03 and they could influence one another or even feedback
00:13:50.08 to influence the first transcription Factor
00:13:52.23 and if you took this model seriously
00:13:56.24 then every one of these arrows is associated with that whole set of parameters
00:14:01.09 and so what we need to know is what is it about the setting of these parameters
00:14:06.12 that allows this whole network to do its job
00:14:08.22 That's why people are exploring many different ideas that would be candidates for a real theory
00:14:14.18 that would explain why these parameters are chosen the way they are
00:14:18.08 by evolution, by physicological adaptation
00:14:20.27 so one idea is that what you want to do is to be sure
00:14:24.09 that the expression levels of these chains carry as much information as possible
00:14:28.08 about what the hell cares about
00:14:29.27 whether that is in the control of metabolism
00:14:32.23 and you're worried about the concentration of various nutrients in the environment
00:14:36.21 and need to capture information about that in order to be sure that you're doing the right thing
00:14:40.16 or you're in a developing embryo
00:14:42.04 and you want to capture information about where you are in the embryo
00:14:45.03 in relation to your neighbors so that you can adopt the correct fate
00:14:48.17 in relationship to the rest of the body plan
00:14:50.18 What's great is that this question about how information flows through a network
00:14:54.21 is something you can think about not just at the molecular level
00:14:58.03 inside a single cell you can think about it and a whole population of organisms
00:15:01.24 you can think about a flock of birds all agreeing to fly in one direction
00:15:06.06 what interactions does he have to have with each other
00:15:08.09 and how did those needs to be tuned
00:15:09.26 in order to be sure that on the one hand that all degree to fly
00:15:12.18 in the same direction efficiently
00:15:14.05 but the other hand they can be
00:15:15.17 reliably responding to stimuli from the environment like the approach of a predator
00:15:21.06 And at an intermediate level you can think about this and networks of neurons
00:15:24.19 where you have the same kinds of input-output relations
00:15:27.25 for the individual neurons, what lives on the arrows
00:15:30.10 are not these molecular processes of the control of gene expression
00:15:33.26 but rather the synapses are connections between neurons
00:15:36.20 and the strength of those connections the parameters if you will
00:15:40.09 along these curves are themselves changing during the course of learning
00:15:44.04 and we can ask what is it that the network is learning to do
00:15:47.04 by adjusting these parameters
00:15:48.22 what's the bigger theory that tells you how all this fits together
00:15:52.12 So we don't have any theories yet we have candidates for them
00:15:55.25 that lots of people are trying out
00:15:57.21 and the whole process is being driven by the fact that
00:16:00.04 our experimental friends can now make these incredibly precise measurements
00:16:04.02 and the result is that we are at a moment
00:16:07.24 where the whole theory in the sense that we use that term in physics
00:16:11.18 the sense of a powerful and often quite abstract mathematical description
00:16:18.04 that can reach down and predict he details of experiments
00:16:22.08 with great quantitative precision
00:16:26.02 that hope for a theory is now coming we can turn into a real research program
00:16:30.04 rather than being the fantasy it was
00:16:32.08 in previous generations. So, I'll stop there. Thanks.

This Talk

Speaker: William Bialek

Audience:

Educators of H. School / Intro Undergrad
Student
Educators of Adv. Undergrad / Grad
Researcher
Educators

Recorded: December 2014

Talk Overview

As biology becomes increasingly quantifiable, William Bialek posits that scientists can develop unifying theories for biology that predict precisely how living systems work. Unifying theories are more powerful than mathematical models because they can be applied to diverse biological phenomena, such as transcriptional control of gene expression and neural signaling. However, as Bialek points out, many challenges lie ahead in constructing these comprehensive mathematical descriptions that define the unifying theories for biology.

Speaker Bio

William Bialek

William Bialek is the John Archibald Wheeler/Battelle Professor in Physics and a member of the Lewis–Sigler Institute for Integrative Genomics at Princeton University. A theoretical biophysicist, Bialek has explored research interests that span the dynamics of individual biological molecules to learning and cognition. He is best known for contributing to our understanding of coding and… Continue Reading