Age of Genomics: The Genome as the Modern Rosetta Stone

Top

00:00:06.18 Good morning, my name is Rob Phillips
00:00:08.17 and I'm from Caltech. Today, I'm at Woods Hole in
00:00:11.29 Massachusetts. Super privileged to be here. And
00:00:15.19 I'm going to be talking about great questions in
00:00:17.20 the life sciences. And these are questions where it's
00:00:21.17 thought that maybe physics and computation will be able
00:00:24.06 to shed some interesting light. Now the question I'm posing
00:00:28.02 today is, can we find a way to do what I would call,
00:00:32.22 decipher the modern Rosetta Stone? Here are just
00:00:37.07 to give you a sense, a historically inaccurate sense,
00:00:40.13 of the rosetta stone and its history. The basic story goes
00:00:44.21 something like this. Napoleon had grand ambitions, he decided
00:00:48.21 to go to Egypt, and when he got there, his soldiers they
00:00:52.01 ransacked the place. And one of the things that happened is
00:00:55.06 they found this rock, the Rosetta Stone. It led to a variety
00:00:59.24 of geniuses, such as Thomas Young, known for his study of
00:01:03.21 interference of light. And Champollion to decipher
00:01:08.07 the Rosetta Stone. Of course, from a historical, from an
00:01:12.02 archeological perspective, this is clearly a very interesting
00:01:14.15 thing. But, I would say that at the moment, we're living through
00:01:18.11 an era where there are much bigger fish to fry. There's
00:01:21.22 much more sort of demanding set of questions of
00:01:24.18 deciphering something. And that's that we're living in
00:01:27.02 the genomic era. So, the question that I'm going to pose
00:01:30.11 today has to do with, okay, we can now more or less
00:01:34.20 at will, sequence genomes. We can sequence genomes
00:01:37.25 of human beings and a generic organism that we discover.
00:01:40.28 Bacteria, whatever. What this leads to is long strings
00:01:45.00 of letters, A's, T's, G's, and C's. But it does leave us
00:01:49.01 wondering, what does it all mean? Now the thing I'm
00:01:52.18 try to do in this talk is to give a sense of the extent to which
00:01:55.25 we successfully had a translation, like that that I showed you
00:01:58.25 here. Which is, it is a sort of translation table for the Rosetta
00:02:03.26 Stone. And the extent to which there are gaping holes
00:02:07.20 in our understanding of this information content of genomes.
00:02:10.28 So, to go back to the 1960s, I want to tell you a little bit
00:02:15.18 about what I think is one of the greatest experiments that was
00:02:18.11 ever done. In any science. This was done by Matthaei
00:02:21.18 and Nirenberg, and it has to do with cracking the genetic code.
00:02:24.13 The idea is that after the elucidation of the structure of
00:02:28.22 DNA by Watson and Crick, so 1953, there was a ten year
00:02:33.18 rush, 15 year rush, to figure out what does the sequence
00:02:37.01 of letters A, T, C, and G signify? And the reason that this is
00:02:41.10 as I would say interesting in a certain puzzle is that we
00:02:44.15 know that there are four nucleotides, A, T, G, and C, as I
00:02:49.13 already mentioned. And we also know that a typical
00:02:52.19 protein has on the order of 300 amino acids. Now if it takes
00:02:57.08 3 nucleotides to code for each amino acid, then that means
00:03:00.29 a typical sequence, a nucleotide sequence, is around
00:03:03.11 4^1000 different possibilities. It's a huge combinatorial
00:03:08.13 space. And the question they set themselves was,
00:03:11.10 how do we tame that huge complexity and beat it down
00:03:14.21 into a form where we would actually have some sort of
00:03:17.21 translation table? The way this experiment happened
00:03:20.05 was synthetic biology, if you like, it wasn't called that
00:03:23.22 at the time. But what they did was they synthesized mRNAs
00:03:27.04 that were repetitive, U-U-U-U-U and so on. And then they
00:03:32.19 fed these mRNAs to a cell extract and asked the question,
00:03:36.24 which of the radiolabeled amino acids are incorporated in
00:03:41.16 the protein? And what they found is a chain,
00:03:43.21 phenylalanine-phenylalanine-phenylalanine-phenylalanine
00:03:46.06 and so on. And the consequence was that they cracked the
00:03:50.06 genetic code. By doing this repeatedly, they were able to
00:03:53.21 essentially figure out the entire genetic code. So a way to
00:03:57.21 think of this, most people have seen this before, you start
00:04:00.19 from the inner circle, that tells you the first codon, the first
00:04:03.17 letter, then you go to the next circle out, that tells you the
00:04:06.08 second one, the third circle out tells you the third,
00:04:09.03 and if you look at that triplet and then that will tell you
00:04:12.06 on the fourth ring, which amino acid goes with that collection
00:04:16.14 of nucleotides. The reason I say in the title that this is only a
00:04:20.21 partial vocabulary, is that it is in a certain sense deceptive.
00:04:24.28 It makes it sound as though we've cracked this Rosetta Stone,
00:04:28.11 when in fact, we really haven't. The reason I say that is
00:04:32.11 illustrated I think, even in the case of the supposedly best
00:04:35.24 understood organism, which is the bacterium, E. coli.
00:04:38.28 If you go look at the databases of what we know about
00:04:42.17 E. coli, what you'll find is that 2000 of the roughly 4500
00:04:46.29 genes have no regulatory annotation. Now what do I mean
00:04:51.05 by that? What I mean is that, we indeed know where genes
00:04:54.28 start. We indeed know that certain amino acids will be
00:04:58.22 reduced and strung together as a result of that gene. But what we
00:05:03.07 know nothing about is the decoration of the DNA in front of that
00:05:07.14 gene that tells it whether it should be on or not, in the first place.
00:05:10.13 In other words, it's regulated. And we really don't know how
00:05:12.22 that works. In the case of eukaryotic cells, animal cells,
00:05:17.06 the situation is even worse, as shown by the lower diagram
00:05:19.29 here. What that is, is it's showing a hunk of genomic DNA
00:05:24.19 and this green part is known as an enhancer. Enhancer refers
00:05:28.27 to a little chunk of DNA that can be as far away as megabases
00:05:33.10 from the gene it controls. And then it bends over and kisses
00:05:37.12 its regulatory region onto the basal transcription apparatus
00:05:41.25 and turns the gene on. The problem is, if I handed you some
00:05:45.19 genomic sequence and said, okay, you tell me where the
00:05:48.02 enhancers are and you tell me which genes they interact
00:05:50.29 with, it would be very hard at best. Okay, so I claim that
00:05:55.20 this means we really haven't cracked the full genetic
00:05:58.01 code. And so, this is to me really the great question of
00:06:01.26 this particular talk. But I want to tell you a little bit about
00:06:05.11 approaches that have been taken. This is from an experiment
00:06:07.22 that was done by Justin Kinney and his PhD thesis work, which I
00:06:11.21 think is really quite cool as a way of looking forward to
00:06:16.04 own the genome with single base-pair resolution. So I show
00:06:19.06 in this diagram, I have some promoter, some region of DNA
00:06:23.14 that controls the gene that I'm interested in. I want to figure
00:06:26.02 -- and I don't know what the regulatory behavior is,
00:06:28.08 I want to find that out. So how do we do that? His idea was,
00:06:31.06 let's take the DNA and mutate it at about 10%, in other words,
00:06:36.08 10 out of every 100 bases will screw up. We'll change them
00:06:39.12 from A to T, and A to G, and so on. And then we'll ask the
00:06:43.01 question, if we wire up cells, each cell having a different version
00:06:48.25 of this mutagenized promoter, how much fluorescence does this
00:06:52.12 cell produce? The fluorescence is used as a way to read
00:06:54.21 out how much we've messed up the gene. And so as shown in this
00:06:59.12 diagram, he makes a library of what are called plasmids, these
00:07:02.10 are small circular DNA with his gene of interest. Mutagenized
00:07:05.21 gene, the cells are fluorescent. Then what he does is
00:07:09.13 sorts the cells according to how much fluorescence they have, that's
00:07:12.19 what you see here, the sorting scheme. And then categorizes
00:07:16.20 them according to how high or low their fluorescence is
00:07:19.22 and that's the point of this bottom diagram. The group called
00:07:22.16 B1, B2, B3, and so on, those are bins in which the cells
00:07:27.07 have different levels of fluorescence. And each collection
00:07:30.13 has a different fluorescence. So that's the key idea.
00:07:33.12 For me, where physics and sort of theory meets this kind of
00:07:38.03 experiment is this graph shown here in the upper right.
00:07:41.01 Which tells us the relationship between sequence and
00:07:44.03 which bin a given cell got sorted into. Let me try and give you
00:07:47.14 an analogy. I'm in a room right now, there's a door, if someone
00:07:51.28 were to walk in the door right now, I could make a guess
00:07:54.08 about the height of that person. How might I do that? Well
00:07:56.21 I would just reason, well, typical human beings are between
00:07:59.10 5' and 6' feet tall, I would do something along those lines.
00:08:02.11 If somebody told me the next person to walk in is going to be
00:08:05.18 a woman, then I would do much better on localizing that person's
00:08:09.01 height. Now why is that? It's because there's what's called a
00:08:12.18 mutual information between height and gender. Just like in this
00:08:17.10 case, there's a mutual information between which base,
00:08:21.05 is it an A, a T, a G, or a C, at a given site. And which bin
00:08:26.15 the cells will be sorted into. So, what comes out of this
00:08:30.03 is what's called an information footprint. Ideas that tells
00:08:34.04 you what bases, at least the ones labeled in orange,
00:08:36.26 which bases are the ones that are important for gene expression.
00:08:40.00 And so, what you can do with this method is, you can take -- go to
00:08:43.21 a region of the genome which you're ignorant about, and you can
00:08:47.09 figure out where transcription factors will bind. And you can start
00:08:51.23 to decode this next layer of information that's hidden in the genome.
00:08:55.23 So, let's think about this. What I've told you is that there's a challenge,
00:09:00.18 first of all, we really don't understand the genome in detail at the
00:09:04.21 -- at all levels of information. But let's say we solve the problem
00:09:08.19 in the -- using the type of scheme that I just talked about.
00:09:11.25 Then what I would say is, we have a translation table
00:09:15.13 like this picture that I showed you earlier, which comes from Champollion,
00:09:19.03 which was basically telling us the meaning of the symbols
00:09:22.04 in the hieroglyphics. But what I want you to consider is, having
00:09:26.17 the list of symbols, if I knew the letters in the English alphabet,
00:09:29.24 and even perhaps, if I knew the meaning of, if I had words in
00:09:32.18 the English alphabet, that doesn't imply that I'm ready to write
00:09:35.11 a novel. So the vision that I really want to finish with is,
00:09:38.21 let's say that one solves this problem of figuring out what the
00:09:41.06 genome says at the level of the words, how do we compose
00:09:44.27 a genomic novel? That's the question that I want to think about.
00:09:47.22 So, in order to argue this, I want to go back in a way to
00:09:52.06 engineering in the physics setting. I think of engineering
00:09:54.27 as synthetic physics, and the reason I put the word "synthetic" in there
00:09:58.02 is because I want to conjure up images of synthetic biology.
00:10:02.01 So, engineering is synthetic physics. And often what one
00:10:06.09 does is the simple problem first, a toy problem. You don't work on
00:10:10.11 the airplane wing first, you work on flow past a sphere or flow past
00:10:14.17 a cylinder, and you work out the rules of that problem and then
00:10:18.10 once you have those rules in hand, then you turn to the more
00:10:21.24 complicated situation. So the beauty of the picture I'm showing
00:10:25.02 above is, this is the kind of thing that Leonardo Da Vinci was already
00:10:28.07 thinking about. He was curious about flow past a sphere, he was able
00:10:32.11 to draw the vortices that you see in the picture. And my point is that
00:10:36.11 after 150 years of fluid mechanics, we started understanding
00:10:40.05 things well enough to then say, let's try to use these principles
00:10:44.11 for something more tricky, which is for example, the shape of
00:10:47.18 an airfoil. I love this picture, which shows, it's a Boeing picture,
00:10:51.16 which shows a 747 and a 787. Really gives you a sense of
00:10:55.26 where we've come with engineering. And the dream I guess
00:11:00.20 I would like to articulate for biology is to have a biological
00:11:03.26 engineering that bears the same relationship to fundamental
00:11:07.10 science that this kind of thing in mechanical and electrical
00:11:11.10 engineering do to physics. So, in a certain sense, what I would like
00:11:18.01 is I'd like to be able to write the novel of the anglerfish. So the idea
00:11:21.15 is, I could go grab a fish, sequence its genome, be able to
00:11:25.16 put the sequence into a computer to read it and say, "Okay, those
00:11:30.08 are the genes, these are the regulatory regions, these over here
00:11:34.08 are, in a certain sense, DNA fossils. And now I will reprogram
00:11:39.29 this fish so that the color of the light that comes out of its little
00:11:43.23 thing on its head is different." That's a vision for synthetic biology
00:11:48.09 that would really be based on the infrastructure in deciphering
00:11:53.16 the genomic Rosetta Stone. In summary, what I've tried to say,
00:11:57.12 I think is a great question of our time, is the search to decipher
00:12:01.14 these genomic Rosetta Stones. We've already learned a lot,
00:12:04.11 but I think that in a certain sense, Thomas Paine once said,
00:12:07.27 a long habit of thinking a thing wrong gives it a superficial
00:12:11.12 appearance of being right. A long habit of thinking that we
00:12:14.21 understand genomes and we've cracked the code gives it a
00:12:18.27 superficial appearance of having been solved, but I think
00:12:21.04 really, it hasn't. And so, to me that's one of the great questions.
00:12:25.01 So I close by just saying it's a subjective matter, this issue of what are
00:12:30.11 great questions. But that's the privilege of doing science, is that we get to
00:12:33.23 ask questions that interest us and we're able to go pursue them.
00:12:37.24 Thanks for listening.
00:12:38.28

This Talk

Speaker: Robert Phillips

Audience:

Educators of Adv. Undergrad / Grad
Researcher
Educators

Recorded: July 2014

Talk Overview

Despite living in the age of genomics, Rob Phillips argues that we still haven’t completely cracked the genetic code. Excluding genes, there are many important DNA elements, such as regulatory regions, for which we know little about their function. Deciphering all of the functional elements of the genome will help scientists understand how life is engineered, allowing them to make huge leaps in synthetic biology research.

Speaker Bio

Robert Phillips

Rob Phillips is the Fred and Nancy Morris Professor of Biophysics and Biology at the California Institute of Technology and co-Director of the Physiology summer course at the Marine Biology Laboratory. His research focuses on elucidating the physical phenomena of a variety of biological processes/systems, such as mechanosensation and gene regulation. He combines physical modeling… Continue Reading