 I'm Magali Haas, I'm the CEO and founder of Orion Bio Networks, which is a nonprofit enterprise and today we're going to be sharing with you our experiences with this effort which launched in January of this year. It's really a privilege to be here with you all today because a lot of the things we're going to talk about the themes are issues and topics that are very familiar to you, but when we go out to the rest of the world and talk about these issues and try to gain funding, gain support for these approaches, we're finding that they're really quite a lot of obstacles and challenges still ahead of us. So I'd like for you to think about what we're going to share with you today as a pilot approach and also as a representative example of where we are today in the external world and the environment outside of INCF. To briefly cover the agenda, we have several speakers today. Orion Bio Networks is an alliance of partners and so I have several partners with me today presenting on everything from clinical to computational topics. I'm going to actually also cover not only my own topic but Dr. Phil De Yeager who's from Brigham Women's Hospital was unable to join us here today due to a family emergency. So why did we start Orion Bio Networks? Well, first of all, I'm a neuroscientist and I'm also a clinician and looking at this world of brain disorders through both the lens of a physician and as a clinical researcher, I would say that our paradigm is broken and it's broken when I look at the system from the perspective of outcomes. We are challenged in classifying brain conditions. We know from the recent DSM-4, DSM-5 controversy that our nosology is faulty and it demands attention. There are over 300 brain disorders, in fact, I can't even tell you what the real count is. I've looked at the NIH sites, the NIMH sites, the ICD sites and the total count is actually not quantified. So how many brain disorders are there? And then more importantly, what are their mechanisms? Why are they still called disorders and not diseases? In order for them to be a disease, we need to know what the mechanism of action was. We have very poor or no definitive diagnostics for most of these conditions so as a physician trying to identify whether a patient has a condition or not, I do not have good objective measures. It's a very subjective way of diagnosing patients. Treatment decisions are not individualized and sometimes arbitrary. Most therapeutics are for symptomatic treatment and we do not know who will respond, whether they will respond and to what extent they will respond so the results are unpredictable. And we do not know what the disease course looks like so we cannot inform our patients what's going to happen to you over the course of your lifetime. This is a problem. So as a clinical researcher, I'm looking at the world and I think we need a new paradigm for brain research and how much are we willing to invest in this remains to be seen. What are some of the challenges? Well I think one of the fundamental challenges and I think this is a parable that you've probably heard many times here at INCF, Sean and I share a very strong visual alignment here. And that is that the parable of the elephant and the blind men where we think of the individual researchers looking at different aspects of the elephant and understanding the elephant in its either based on its appearance in terms of the tusk, the tail, the skin and how you interpret and how you understand the disease is entirely dependent on which component of the elephant you have to be evaluating. And unless you have a mechanism by which you can take all that information and integrate it, you never get to see the bigger picture, you never get to see the whole elephant. Furthermore, if our funding model is set up so that we're only funding subsets of the elephant to be studied, so everybody's looking over here and no one's ever looking over here, that's also a challenge. So I think that part of the paradigm shift that could be occurring in our research enterprise is to think about how we can use disease modeling to integrate knowledge and understanding and model diseases instead. And so when I went to funders originally with this idea for Ryan Bionetworks, the question was why aren't we investing in computational modeling as an approach to help us identify new therapeutics, diagnostics and cures for disease? And the answer was because it's not ready yet, or it's so we think. But then I asked them, how do you make decisions? How do you decide where to invest limited capital and resources? These are real issues in the business world. And the answer is, well, we have to have a certain sense of confidence, right? If we're going to go down a particular path, a certain course of action, you have to have confidence that whatever action we're going to take is going to have the outcome we intend. Even though we are failing, most of our trials fail in phase three, we still go down that course to get to that failing point. So in order to improve development, in order to improve clinical decision making, we need to have better confidence intervals. We need to have a better predictivity of disease and outcomes. In other words, prediction is probabilistic. It's not yes or no. The answer is yes. It's not Boolean. It's not a simple switch. It's probabilistic. What is the probability that this molecule will hit this target, that it will have a beneficial effect on this disease? And we will have results at the end. So probabilistic thinking and Bayesian probabilities, something that has been around for over 200 years. So this is not new. And the mathematics is well established and continues to evolve very rapidly. So applying this disease system biology is something that has only really been considered more recently when we saw the advancement of supercomputing. And again, these are all things that this audience is very familiar with. What you're not familiar with is that even though you have the math and you have the supercomputers, you don't have the ecosystem that's going to actually allow this to advance disease. And so this is where the concept of Orion's network emerges. The idea behind it is how do we get the right data? How do we get big data? How do we get the big analytics to support that, the computational environment, the capabilities, the algorithms, all together? And how do we get the right people together in a room with the right questions? What are the questions we're trying to answer around a given disease? Do we even understand what we don't know? Do we know what we know? Do we know what we don't know? How do we get the right data to answer those questions? Very basic things. But the question is, how do we create the ecosystem to do this? And this is the idea behind Orion Bionetworks, is to get a cooperative alliance of partners with different capabilities around the table and provide data, expertise, technical knowledge, computational knowledge, and funding to allow this to happen. So let's start with the data. We need a lot of data to understand disease. And frankly, in most of brain disorder research, we've focused either here at the syndromic level, clinical phenotypes, or we have focused here more recently with the genetics, GWAS copy number work that's being done. There's very little research being done in the middle. So if we want to bridge this divide between genetics and syndrome, we need a lot of data about what's happening downstream of those genes, their functional consequences, their impact on signaling pathways, the connectome, and then their impact on behavior, whose brain disorders manifest largely as behavioral disorders. And how do we quantify those effectively with the phenotype, which Jamie and others will talk about? And then, of course, it's gene by environment. So we need to understand what's going on in your environment. Your friendly neighborhood microbiome that's with you every day, and how does that influence you? The environmental factors at birth throughout your life that add aggregate impact. In a disorder like traumatic brain injury, the number of traumas to your head will increase or decrease the likelihood of you having a bad outcome over your lifetime. Your own internal state of mind, and then, of course, therapeutic interventions and the way they interact. So what we want to build is this human-interactile model. But most of the data that we have in hand today, if you think about large cohorts, is here and here, not so much here, and not so much there. So we need the right data sets. Then we need a way to assemble those and start building computational models. And what we want to think about is, what are those models? Now, there are a lot of organizations out there that do what I would call, and I am not a specialist in this area, bottom-up modeling. So you're building up models based on the disease biology, what you know. But I just told you we know less than 3% of the biology. So how are you building up those models? That's where I think some of the causal inference modeling and hypothesis-independent modeling approaches comes into play. This is a way that we can look at all of these large data sets and start to streamline what's relevant, what variables predict the outcomes we're interested in, cause and effect relationships, not just correlations. And then drive that through to confirmatory biological experiments, start to create a list of priors that then do build these predictive, hierarchical disease models that people like Stephen Larson later on will tell you about. In an iterative fashion, then, we can start creating an environment where we collect the right quantitative data, we model it, we identify what we think we know, we start to model that further, and we enrich the data set itself over time. And finally, the cooperative alliance. Who do you bring together around this table? Interdisciplinary teams. What we've constructed here for Orion 1.0 is a multidisciplinary team of partners that provide data, high quality data from the Brigham Women's Hospital in Boston from the Accelerated Cure Project, which is a non-profit enterprise. I think this organization spearheaded the idea of creating biomarker repositories, and we will hear shortly from Robert McBurney, who leads that organization. And then we need the system biology partners, and Orion has several. So we have GNS Healthcare, Metacell, Thompson Reuters, and also Rancho Biosciences, all working on the same data sets, and evaluating them with different methodologies, so that we can determine which of these actually provide predictive potential and high confidence in the results. We also need funding, and that comes from the government, and that comes from pharmaceutical partners. This is our Alliance leadership team. You'll see these familiar faces today. One face you will not see today is that of Filda Yeager, who is one of our MS experts. And the data sets themselves, and this is one of the first lessons we learned in this exercise, we started with three major data sets that we thought would inform the exercise, the 2,500 patients coming from the Brigham Women's cohort, this is a longitudinal study of patients with MS. Most of them coming into the clinic with newly diagnosed multiple sclerosis. And a variety of measures evaluated in these patient population, including demographic clinical imaging treatment, and also genomics. As it turned out, because of the way that clinical research is conducted today, these assessments were not done in a coherent fashion. And so at the end of the day, when we start to look at the data, the number of patients that we could actually bring into the computational modeling exercise was 108, meeting all the criteria that we wanted to meet at the outset. So from 2,500 data sets we got down to 1,108. And then we have the accelerated cure data set, and I'll let Robert tell you more about that. And then the patients like me data set, which gives us an entirely different view on phenotype from the patient's perspective, and Jamie will share more on that. So the purpose of Orion 1.0 was first, can we get the right data sets together? Can we bring them all into one integrated database? Will different institutions share their data? Secondly, can we start building prognostic models of MS, computational methods? Can we share those broadly across the alliance? And does this give us enough of a belief that this approach is viable, that we can go further and start developing iterative solutions from here? So hopefully by the end of this session today, you'll see that Orion 1.0, which is now on its seventh month and almost completed, has provided a lot of answers to those questions. And we are embarking on now Orion 2.0 activity heading into 2014, where we plan to scale up this effort with additional multimodal data being collected on a prospective basis, now informed by the knowledge of what kind of data you need and why you need it in a certain way. New models and new modeling approaches, new partners from the computational realm. An expanded toolbox of bioinformatics simulation and data tools, which Stephen will tell you more. And the ability then to work with basic biology partners to do validation of some of the findings that we are seeing already. And most importantly, dissemination and education so that we can build this field over time and hopefully embrace this approach for the future. Neuroinformatics is going to be the basis of disease research. It's unquestionable. The question is, how do we get there systematically? So that's my introduction to why we built Orion. Now I'm going to change hats. I'm now Filda Yeager. I'm the head of the MS Institute of Neuroscience at the Brigham Women's Hospital. I've spent my life working as a physician in MS and I have an active clinical practice. I'm an MD-PhD and I do research at least half my time. And I work with Howard Weiner who heads up the Brigham Women's Hospital entire MS program and spearheaded a very important initiative called the Climb Study over 10 years ago, starting this longitudinal cohort to be collected in patients with MS. And MS is a very complex disease. It's a neurologic disease, at least that's our current nozology, with both inflammatory and neurodegenerative components. And 40% of patients have comorbid depression. It is a fairly prevalent disease, one in 1,000 individuals of predominantly European descent. It affects women disproportionately over men. And most of the therapeutic targets today address the immune cell, what we understand of the etiology of MS. So let me tell you a little bit about what we think we know about the natural history of MS and then what we don't know. First, natural history of MS is a remitting disorder. So you start out with initial symptoms and this is a relapse event. And over time, the relapses without any other intervention keep coming back. And you can also see that there's a steady decline or a neurodegenerative process that overlaps that. And that can be quantified as a reduction in your overall brain volume over time. Whether or not these relapses are one in the same etiology as this neurodegenerative process remains to be determined. We don't know. We know that we can impact these through therapeutic interventions and ones that are available today, but have no impact on this neurodegenerative process as we go. And what you see here are MRI images showing that the decrease in the brain volume, and this is in the secondary progressive population once you've gotten past this stage. This is the period that we call the MS susceptibility period. And we know that there are significant genetic factors that increase your likelihood of developing multiple sclerosis. And these are some of the agents that are predominantly used for the relapse remitting component of the disease. You'll notice that we have no treatments for the neurodegenerative component. There has been an international effort led by the Brigham Women's Hospital Group and including all of these different organizations, one of these very large-scale efforts to look at the genetics of MS. And here are some of the initial findings. Initially, when we had a limited number of data sets, we found two genes associated with the disease. And as this been typical of the genetics field, the higher the number of cases and controls, the higher the number of now validated genes that we see affecting the disease. So we now know that over 100 new genes have been identified associated with the risk for developing MS. And what we don't know is, what those genes do? What is their downstream function? How do they impact the disease course and why? We also know that the disease is incredibly heterogeneous in terms of its disease course. So here we have two MRI images, one of a 29-year-old woman with MS and one with a 34-year-old woman with MS. And you can see here some of the lesions. These are the plaques that are typical of the disease. This woman has a fairly normal volume, brain volume overall, but you do see a plaque here. And you can see this brain. If I hadn't told you this was a 34-year-old woman, you would have thought, you could have thought this was a 65-year-old woman. This is the brain of someone with significant neurodegeneration and looks like an Alzheimer's brain. Yet she's 34. So how do we deal as clinicians with this incredible heterogeneity? How do we know who we should offer what therapeutic? And some of the experiments that we've done in the lab at Brigham, peripheral blood RNA profiling shows that you can define different subcohorts, stratify the population, but it's very difficult to interpret this kind of data absent any other input. So at the end of the day, what we really need is an integrative approach to uncover the population structure and start to distill the genetics, the MRI and the blood profile and bring that all together to determine where patients really, how they really stratify. And frankly, in our lab, our approaches have all have not taken advantage of computational approaches. So we were fascinated by the idea of working with partners like GNS, Metasol and others on this project. We have many unmet needs in MS. We need to structure the patient population. Like I said, by stratified approaches, we have to understand brain atrophy better, and we need primary prevention, but we don't know what targets to pursue. Our analyses to date have been either one-dimensional analyses of biomarker differences or something like the genetic GWAS consortium approaches. We have yet to really apply in this field, and this is one of the leading centers in MS, these types of computational approaches. So Orion was one of the first opportunities to do that. And we're going to see the results of this later. So that ends Phil DeYager's talk. And I'm now going to transition to Jamie Maywood. And we're going to take questions all together at the end.