WEBVTT 00:00:00.000 --> 00:00:26.000 [MUSIC] 00:00:26.000 --> 00:00:31.000 So, in many ways a broken car is not so different from a disease, 00:00:31.000 --> 00:00:34.000 when the engine is smoking and the lights don't come up. 00:00:34.000 --> 00:00:38.000 There's a fundamental difference, however, between humans and cars. 00:00:38.000 --> 00:00:44.000 If I can get my car to a mechanic, I can be pretty certain that they can fix it, 00:00:44.000 --> 00:00:49.000 which is much more than we can say about many of our diseases today. 00:00:49.000 --> 00:00:56.000 So why can a mechanic, with much less education and much less bucks than a doctor, 00:00:56.000 --> 00:01:02.000 fix our car, while our doctors often let us go with diseases persisting in our body? 00:01:02.000 --> 00:01:08.000 Well, there are a number of things that actually a mechanic has that our doctor doesn't have right now. 00:01:08.000 --> 00:01:12.000 First of all, it's got a parts list. 00:01:12.000 --> 00:01:16.000 It has a blueprint telling us how the pieces connect together. 00:01:16.000 --> 00:01:21.000 It has diagnostics tools to figure out where the components, which is broken and what is healthy. 00:01:21.000 --> 00:01:25.000 It has the means, essentially, to replace the parts. 00:01:25.000 --> 00:01:29.000 Now let's think about it. Which of these components are available to our doctor today? 00:01:29.000 --> 00:01:34.000 Well, the good news is that they've finally got the parts list. 00:01:34.000 --> 00:01:38.000 That was the outcome of the Human Genome Project. 00:01:38.000 --> 00:01:42.000 And when the human genome was actually mapped about ten years ago, we thought 00:01:42.000 --> 00:01:48.000 It's going to be easy from now. From the parts, we will have essentially the world bonanza that we need to fix us, humans. 00:01:48.000 --> 00:01:58.000 But of course reality sinks in. We also thought that these many pieces will eventually give us lots of drugs. 00:01:58.000 --> 00:02:05.000 In 2001, or 2000, the year before the genome project was unveiled, the FDA approved about a hundred drugs per year. 00:02:05.000 --> 00:02:09.000 We thought this number could only go up. 00:02:09.000 --> 00:02:12.000 It could only just increase. 00:02:12.000 --> 00:02:14.000 Yet the reality sinks in. 00:02:14.000 --> 00:02:19.000 The number of new drugs in just the last ten years, went from a hundred before the genome, to about twenty per year. 00:02:19.000 --> 00:02:24.000 In hindsight, the reason is pretty clear. 00:02:24.000 --> 00:02:28.000 It's not enough to have the parts list. 00:02:28.000 --> 00:02:32.000 We also need to actually figure out how the pieces fit together. 00:02:32.000 --> 00:02:40.000 That is, we should not look at this picture, but rather we should be looking at how the wiring diagram of the car looks like. 00:02:40.000 --> 00:02:44.000 How the wiring of ourselves actually look like. 00:02:44.000 --> 00:02:51.000 How the genes and the proteins and the metabolites link to each other, forming a conistent network. 00:02:51.000 --> 00:02:58.000 Because this network, which I am going to try to tell you today, is really the key to understanding human diseases. 00:02:58.000 --> 00:03:05.000 Now, the problem is that if you look at this map, you soon realize that it looks completely random. 00:03:05.000 --> 00:03:08.000 Randomness certainly has the upper hand. 00:03:08.000 --> 00:03:13.000 But down the line, it is not. I believe there is a deep order behind this wiring diagram. 00:03:13.000 --> 00:03:19.000 And understanding that order is the key to understand human diseases. 00:03:19.000 --> 00:03:25.000 Now, I am a physicist, and the conventional wisdom is that as a physicist, I should be studying very big objects: 00:03:25.000 --> 00:03:29.000 stars, quasars, or very tiny ones like the Higgs boson and quarks. 00:03:29.000 --> 00:03:36.000 Yet about a decade ago, my interest has turned to a completely different subject: Complex systems and networks. 00:03:36.000 --> 00:03:44.000 And that's because our very existence depends on the successful functioning of systems and networks behind us. 00:03:44.000 --> 00:03:54.000 And I also believe the scientific challenges behind complex systems and networks are just as deep as behind quarks or quasars. 00:03:54.000 --> 00:03:58.000 So I started looking at the structure of the Internet. 00:03:58.000 --> 00:04:04.000 Telling us how many, many computers are linked together by various cables. 00:04:04.000 --> 00:04:14.000 I looked at the structure of the social network, telling us how do societies wire together through many friendship and other linkages. 00:04:14.000 --> 00:04:18.000 And eventually I started looking at the structure of the cell. 00:04:18.000 --> 00:04:23.000 Telling us you our genes and proteins link to each other into a coherent network. 00:04:23.000 --> 00:04:27.000 And through that path, I arrived at human diseases. 00:04:27.000 --> 00:04:30.000 A path that is rarely taken by physicists. 00:04:30.000 --> 00:04:35.000 Now, the fundamental question that really comes up from that is: 00:04:35.000 --> 00:04:43.000 How do we think about diseases in the context of these of these very very complicated networks? 00:04:43.000 --> 00:04:51.000 And from that, let me turn to a map that we all understand, probably the most famous map out there, which is the map of Manhattan. 00:04:51.000 --> 00:04:57.000 Now, in many ways, Manhattan is structured different from a cell. 00:04:57.000 --> 00:05:04.000 But let's for a moment carry with me and let's believe together that this is really not a map of Manhattan but a map of a cell. 00:05:04.000 --> 00:05:09.000 Where the intersections showing us nodes are the genes and the proteins. 00:05:09.000 --> 00:05:14.000 And the street segments that connect them corresponds to the interactions between them. 00:05:14.000 --> 00:05:20.000 Now, down the line, this is not so different from what happens in our cells. 00:05:20.000 --> 00:05:27.000 The busy life of Manhattan very easily maps into the crowded life of the cell where molecules interact with each other, 00:05:27.000 --> 00:05:30.000 and recombine and transport and so on. 00:05:30.000 --> 00:05:35.000 So there's lots of similarities on the surface between them. 00:05:35.000 --> 00:05:39.000 And if we look at Manhattan, we also realize that action is not uniformly spread within the city. 00:05:39.000 --> 00:05:46.000 If you want to go, for example, to the theater, you don't go to ANY parts of Manhattan, you would go to the theater district. 00:05:46.000 --> 00:05:50.000 Because that's where most of the theaters are, that's where the shows are. 00:05:50.000 --> 00:05:57.000 If you want to buy an artwork. You will not actually be going ANYwhere in the city, but you would be going to the gallery district. 00:05:57.000 --> 00:06:04.000 Because there is one small region in the town that has most of the high-end galleries, and that's where most of the artwork is for sale. 00:06:04.000 --> 00:06:06.000 The same is true in the cell. 00:06:06.000 --> 00:06:10.000 Its functions are not spread uniformly within the network. 00:06:10.000 --> 00:06:14.000 But there are other pockets within the network that are responsible for particular functions, 00:06:14.000 --> 00:06:18.000 and their breakdown potentially leads to disease. 00:06:18.000 --> 00:06:22.000 So the way to think about disease in the context of the network is to think that 00:06:22.000 --> 00:06:26.000 there are different regions that correspond to different diseases on this map. 00:06:26.000 --> 00:06:31.000 So, for example, you could say that cancer stays somewhere around Wall Street 00:06:31.000 --> 00:06:33.000 [AUDIENCE LAUGHTER] 00:06:33.000 --> 00:06:38.000 And bipolar disease would be somewhere around Times Square. 00:06:38.000 --> 00:06:39.000 [AUDIENCE LAUGHTER] 00:06:39.000 --> 00:06:46.000 And you know asthma, a respiratory disease, it would be somewhere up next to the Washington Bridge. 00:06:46.000 --> 00:06:51.000 Where Manhattan breathes its people and cars into New Jersey and The Bronx. 00:06:51.000 --> 00:06:52.000 [AUDIENCE LAUGHTER] 00:06:52.000 --> 00:06:55.000 Now, under normal conditions 00:06:55.000 --> 00:06:58.000 Manhattan is full of traffic. 00:06:58.000 --> 00:07:01.000 And that's how the cell looks like normally. 00:07:01.000 --> 00:07:07.000 But if we had defects, some genes breaking down, that corresponds to some of the intersections not working, and 00:07:07.000 --> 00:07:12.000 soon enough we would get a very typical Manhattan disease: A traffic jam. 00:07:12.000 --> 00:07:16.000 This is not so different from what happens in our cells. 00:07:16.000 --> 00:07:19.000 Because there are many different ways you can get the same phenotype. 00:07:19.000 --> 00:07:24.000 In the same way, there are many different ways you can get a disease. 00:07:24.000 --> 00:07:33.000 For example, there was a famous study by Burt [???]'s group which sequenced about 300 individuals who all had colo-rectal cancer. 00:07:33.000 --> 00:07:34.000 They had the same phenotype. 00:07:34.000 --> 00:07:41.000 Therefore the expectation was that all of them would have probably the same mutations in the same genes. 00:07:41.000 --> 00:07:49.000 Yet, the study showed that not only did they not have the same set of mutations, but the mutations were all in different groups of genes. 00:07:49.000 --> 00:07:56.000 There were no two individuals who would actually have the same genes exactly the same group of genes' defect. 00:07:56.000 --> 00:08:03.000 The only way to understand how it's possible that many different genes broken down in different combinations linked to the same disease, 00:08:03.000 --> 00:08:06.000 is to think in terms of Manhattan. 00:08:06.000 --> 00:08:12.000 If you think in terms of disease module and to really have the wiring diagram of the disease module, 00:08:12.000 --> 00:08:15.000 to understand the breakdown modes of the particular system. 00:08:15.000 --> 00:08:26.000 Now, if we really believe that particular picture, the next step for us is to say, how do we proceed from here? 00:08:26.000 --> 00:08:31.000 It's very easy. Get the map, find the disease module, and drug it. 00:08:31.000 --> 00:08:34.000 Now of course, you do realize there's a catch here. 00:08:34.000 --> 00:08:40.000 And the catch of course is, unlike for Manhattan, we don't have yet a map for the cells. 00:08:40.000 --> 00:08:46.000 I mean, we do, but some of the maps we have are very incomplete. 00:08:46.000 --> 00:08:49.000 For example, the best protein interaction that we have right now, 00:08:49.000 --> 00:08:54.000 we believe it has only five percent of the links that are supposed to be in our cells. 00:08:54.000 --> 00:09:01.000 Now, having five percent of the links means that we are missing 95% of the links. 00:09:01.000 --> 00:09:04.000 And that has dramatic consequences on the system. 00:09:04.000 --> 00:09:07.000 Let me illustrate that on Manhattan. 00:09:07.000 --> 00:09:14.000 Let's go ahead and take 95% of street segments and remove it from the map, and let's see what does it do to Manhattan. 00:09:14.000 --> 00:09:20.000 And the consequence is obvious. The network is broken into tiny pieces. 00:09:20.000 --> 00:09:25.000 And as a result, the modules, the Wall Street neighborhood and the Times Square neighborhood 00:09:25.000 --> 00:09:29.000 that were clearly distinguishable before would be all over the map. 00:09:29.000 --> 00:09:32.000 You don't know any more where your disease module is. 00:09:32.000 --> 00:09:34.000 So what can we do then? 00:09:34.000 --> 00:09:38.000 Well, first and foremost, we must improve on our maps. 00:09:38.000 --> 00:09:44.000 And that's what my colleague Marc Vidal does at Dana-Farber Cancer Institute, 00:09:44.000 --> 00:09:50.000 who in the last twenty years has developed a whole series of automatic tools to systematically map 00:09:50.000 --> 00:09:55.000 the protein interactions within the cell, one of the very important components of the cellular network. 00:09:55.000 --> 00:10:02.000 As a result of his work, a few years ago, we got what we call the 5% map, the one I referred to earlier. 00:10:02.000 --> 00:10:09.000 This year, he's about to unveil another landmark: the 20% map of the human cell. 00:10:09.000 --> 00:10:14.000 And if we left him on the same track, actually he would do the full network. 00:10:14.000 --> 00:10:21.000 It may take a decade or two to get to it, but eventually [???] and many others, we will get a map. 00:10:21.000 --> 00:10:23.000 But what until then? 00:10:23.000 --> 00:10:25.000 Shall we just wait for him to finish the work? 00:10:25.000 --> 00:10:27.000 And the answer is, well, not really. 00:10:27.000 --> 00:10:31.000 There's lots of things we can actually do using the existing maps. 00:10:31.000 --> 00:10:33.000 This is how the map looks like right now. 00:10:33.000 --> 00:10:36.000 This is all the interactions we believe should be in the cell. 00:10:36.000 --> 00:10:41.000 And in order to understand where diseases lie in that, what I'm going to do next is 00:10:41.000 --> 00:10:46.000 I will go ahead and place on this map a particular disease, in this case asthma. 00:10:46.000 --> 00:10:52.000 Asthma is a respiratory disease that leads to coughing, shortness of breath, and many other symptoms, and 00:10:52.000 --> 00:10:57.000 there has been tremendous amount of research on the [???] genetic origins of asthma. 00:10:57.000 --> 00:11:01.000 Therefore, we about a hundred genes that are known to be associated with asthma. 00:11:01.000 --> 00:11:07.000 So if we put them on the map, and I'm showing them now here as purple nodes, 00:11:07.000 --> 00:11:10.000 then we would expect them to be all together. 00:11:10.000 --> 00:11:12.000 Because they really should correspond to our disease model. 00:11:12.000 --> 00:11:14.000 But they're not. They're all over the map. 00:11:14.000 --> 00:11:20.000 And the reason they're all over the map is because we're missing 95% of the interactions. 00:11:20.000 --> 00:11:25.000 So the critical links that would really hold them together in one module are all gone, they are not there yet. 00:11:25.000 --> 00:11:29.000 So what is it we can do next? 00:11:29.000 --> 00:11:31.000 We can use the power of the network. 00:11:31.000 --> 00:11:39.000 They are really built into the network and try to figure out other genes that may also be involved in asthma, about whom we don't know yet. 00:11:39.000 --> 00:11:41.000 And that's exactly what we did next. 00:11:41.000 --> 00:11:48.000 We took this map and we run algorithm through that, that really extract the information from this map, 00:11:48.000 --> 00:11:50.000 and identify what you see in front of your eyes. 00:11:50.000 --> 00:11:53.000 The asthma module within the cell. 00:11:53.000 --> 00:12:00.000 Now if we know the asthma module, from there we can understand the disease's mechanism, the disease's pathways, 00:12:00.000 --> 00:12:04.000 and one day can actually help us understand the drugs. 00:12:04.000 --> 00:12:07.000 But this is not only true for asthma. 00:12:07.000 --> 00:12:10.000 Not only asthma is located well in the network. 00:12:10.000 --> 00:12:15.000 You can take some other diseases, for example COPD, and try to do the same thing. 00:12:15.000 --> 00:12:18.000 COPD is often called the smokers' disease, 00:12:18.000 --> 00:12:20.000 because smokers are at a very high chance of getting it, 00:12:20.000 --> 00:12:23.000 and has somewhat similar symptoms to asthma. 00:12:23.000 --> 00:12:29.000 Not surprisingly, it seems to be that the two modules are significantly overlapping, 00:12:29.000 --> 00:12:32.000 and are certainly in the same region of the network. 00:12:32.000 --> 00:12:38.000 We do expect, however, to have other diseases that would be in a completely different part of the network. 00:12:38.000 --> 00:12:43.000 And what is crucial here is to understand that the relationship between these diseases, 00:12:43.000 --> 00:12:48.000 to what degree they overlap, and how they relate to each other is really crucial to understand 00:12:48.000 --> 00:12:52.000 how they relate to each other. 00:12:52.000 --> 00:12:55.000 and whether they are similar or very different from each other. 00:12:55.000 --> 00:13:01.000 So one way to look at it is to let's look at the relationship of all diseases. 00:13:01.000 --> 00:13:03.000 And that's what I'm showing you here. 00:13:03.000 --> 00:13:09.000 Here in the next slide, every node corresponds to a particular disease, 00:13:09.000 --> 00:13:13.000 and two diseases are connected to each other if they share a gene. 00:13:13.000 --> 00:13:18.000 Why would you do that? Because if they share a gene, then very likely their disease module overlaps, 00:13:18.000 --> 00:13:22.000 and therefore they must be in the same region of the network. 00:13:22.000 --> 00:13:28.000 And what is amazing about this map is that there are links between apparently unrelated diseases, 00:13:28.000 --> 00:13:35.000 which is telling us that if you really want to treat--if you have two diseases and want to treat them, today you may go to different doctors, 00:13:35.000 --> 00:13:37.000 different hospitals, different floors. 00:13:37.000 --> 00:13:41.000 But down at the level of the cell, they are not independent of each other. 00:13:41.000 --> 00:13:45.000 They're connected because they're rooted in somewhat the same neighborhood. 00:13:45.000 --> 00:13:52.000 So what this is telling us, this "diseasedom" as I will call it, is that if we want to understand disease, 00:13:52.000 --> 00:14:01.000 we should not be looking really at what we normally look at, but we should be looking at the network within our cells. 00:14:01.000 --> 00:14:09.000 This is the one that really matters. This is the one that really should tell us how to classify diseases. 00:14:09.000 --> 00:14:10.000 You know, we probably got it fundamentally wrong. 00:14:10.000 --> 00:14:13.000 It's not heart, it's not brains, it's not kidneys. 00:14:13.000 --> 00:14:20.000 Sooner or later we must abandon this organ-based description of disease and turn to what really matter. 00:14:20.000 --> 00:14:31.000 We should stop training cardiologists and neurologists, and rather the doctor of the future needs to become a bit of networkologist, 00:14:31.000 --> 00:14:36.000 to understand where diseases are lying within that network and how they relate to each other. 00:14:36.000 --> 00:14:45.000 So I personally believe we need a new medicine, to truly execute the paradigm change that genomics allowed us to achieve. 00:14:45.000 --> 00:14:52.000 I would call it network medicine, and I think it's really within our footstep to do and achieve that. 00:14:52.000 --> 00:14:59.000 I also think that network medicine will not only help us understand the mechanism of disease, 00:14:59.000 --> 00:15:10.000 but it will affect all aspects of healthcare, from the role of the environment all the way to how we actually deliver care to a particular patient. 00:15:10.000 --> 00:15:23.000 So, coming back to our original question, the good news is that doctors are increasing many of the tools that the car mechanic has today. 00:15:23.000 --> 00:15:27.000 If you think about it, the genomics provides the parts list, 00:15:27.000 --> 00:15:30.000 metabiomics and proteomics provide the diagnostic tools, 00:15:30.000 --> 00:15:38.000 and gene therapy is really giving us the way one day to replace the components, with the pieces that are not broken. 00:15:38.000 --> 00:15:42.000 But a car mechanic would be useless without a blueprint. 00:15:42.000 --> 00:15:49.000 And in the same way I believe that to truly understand diseases, we need to give into the hands of our doctors the map. 00:15:49.000 --> 00:15:56.000 Now I'm a physicist, and a network scientist. I am not a medical doctor. 00:15:56.000 --> 00:16:00.000 Hence, I will never cure any of your diseases. 00:16:00.000 --> 00:16:04.000 I can help, however, decipher the map: 00:16:04.000 --> 00:16:09.000 The real book of life, the book that is currently missing most of its pages. 00:16:09.000 --> 00:16:16.000 But once we learn to read it, we'll get much closer to the secret of life and curing disease. 00:16:16.000 --> 00:16:18.000 Thank you very much. 00:16:18.000 --> 00:16:22.000 [APPLAUSE]