 Hi everyone. I'm an engineer and I was a practice engineer for a number of years and I got involved in medical research because my brother Stephen was diagnosed with motor neuron disease and I then worked in preclinical discovery, ran an institute for a number of years that I founded and started patients like me substantially to try and figure out some of the missing variables that I was working on in ALS. I want to explain a little bit about what patients like me is and then talk a little about some of the work we've done in the context of this network and I think some of the broader implications from that. So what patients like me is is a essentially a dating site crossed with a clinical trial. And what I mean by that is that you come into the site and you can you know plug in variables or things that you're interested in you can find other people like you and you begin to do some rudimentary statistics to understand that and I'll give you a brief demo in a minute, but for the patients they come in and they create a profile usually with a photo. They go through and they find support from others. They enter information about what they're doing and that all becomes an information resource that we use to do research behind the scenes. The system now is about seven years old. It is about 210,000 members. These are our top 20 conditions by numbers and you can see both the patient counts in 5-per-mile. We have 37,000 people. It's the largest registry in the world by a factor of 10. The MS registry is now the largest being anything else by good 15,000. And these numbers you can see sort of the primary and the patient count. What that means is when diseases are comorbid, this is the one that they consider to be the most important condition. And it varies all the way down through panic disorders and psoriasis and a variety of other conditions. There are over 2,000 conditions represented in the system across these 210,000 members. We collect data on surveys that are done often uniquely in custom clinical observations on quality of life, behavior, labs, and then there's forums and free text. We work with pharmaceutical companies, payers, CROs, clinical trial investigators, and researchers in academics. And the result of our research is accomplishing subtyping or signal optimization, which is can you learn what a disease is and how to measure it, disease management and adherence. These are all the things that we work on together as a business. So we're not attempting to solve all of medical research. And I think this is the point of the consortium at some level. You'll see here that if you look at sort of seven, sorry, six variables in disease, disability symptoms, quality of life, lesion burden, inflammation, atrophy, how well can a PRO, a patient reported outcome capture that versus a neurology exam or objective measures or an MRI or autopsy, which is obviously a fairly interventional diagnostic. And these different tools have different strengths. And it's when you combine them as has been executed in this network and you can begin to infer across them that I think you really begin to learn something. Very briefly, I, you know, just show you our MS community snapshot. This was about 2012. You know, this is the number of diagnoses of relapsing remitting, unknown, secondary progressive. And this is a subset of the data that we use to include in this project. So the ones with sufficient data that we felt we could intersect them against others. And I'm just going to very quickly show you what the site looks like from a demo standpoint. And just bring up the, so I can do this, two monitors. Are we seeing this? We are. Look at that. Okay. So this is the site. You sort of join, you can all do it now. I'm not logged in, so you can go and play with it on your own. And if you go and you find the patients tab and look in, you can find patients. So this is just the most recent patients that have updated in the last hour. And you can see there's a fibromyalgia patients. This gets effective disorder, ALS, dysthymia, major depression, and then a few MS patients. So if I could just refilter this just like a dating site and say I'm looking for MS patients. So it's going to convert the 210,000 down to 37 females, say age 35 to 50, again, dating site. And you're automatically re-screening the data very quickly to a new subpopulation. And you'll see, you know, the characteristics, you'll see different patients. Some are very healthy and green, some are very sick. So I could add other variables here where I looked at filtering by how sick you are. So I could say I want MS patients who are only in the MSRS of 40 to 70, which are fairly sick. And it will redraw the data and show you only patients that level of illness. Or how long they've had illness, you know, their diagnosis type or first symptom. If I show you just a quick patient here, this is a woman that sort of given us permission to share her data. She joined PLM in 2007. She's tracked data in the system for over five years. The average MS clinical trial is about two years or 18 months. So this is a longer data set than exists in almost any other registry. You see the quantity of data. These are, I'll show you some of these measures in a minute. But you can see that we collect, you know, about 40 variables at multiple time points. And each of these would tell you how disabled she was at that moment. A Gantt chart of her treatments. And again, it's all very flexible software. So you can just adjust the time scale, get retrospective data. You can see her journey from Capaxone to Avanax through the treatments to Jelenia, which is the latest drug she's launched. And then down here, you see her symptoms and the density of the data about anxiety, depression and other data. So this is sort of typical in the system. This is actually a high dense patient, but this is sort of typical, the kind of data we collect. And this is what we use to generate the information that I'm going to show you. So I'm going to switch now and just come to the outcomes of all of this. So this is our profile. We skew a little female, skew a little younger, but actually are fairly representative. And we're more U.S. than the rest of the world. We're only in English at the moment. That'll change. So why are we doing this? You know, what's the purpose of this whole thing? And I, you know, Magali talked about the broken system. I'm an engineer, and you know, and I know most of you are sort of more mathematicians, but I speak a lot at clinical conferences. And I get very frustrated with medicine's obsession with poetry because I'm much more interested in math. And I think Kelvin really put this very well. The problem with medicine is that when you try and match whatever you're looking at against multiple sclerosis, you're matching some quantitative assessment against poetry. And I think you put it really well, knowledge of a meager and unsatisfactory kind. And if we're to understand the disease, we have to understand it mathematically, express it in numbers. And this is sort of what we're trying to do. So I think what we're trying to do is what is MS from a phenomic level? And our company considers these sort of units fens, it's an internal term we use to sort of not bias ourselves, which are attributes about someone that can be characterized quantitatively. And fens are much more like proteins than genes. They have a much more complicated set of measurement paradigms and components. And if you think about that, these sort of some of the variables we are interested in. So specifically, if we ask about walking, are we being specific? How accurate is it? What's the range? Does it cover the full spectrum of possibilities? How precise is the measure? What are the biases of measuring it in that way? And then we're really interested in these sort of second-order questions, given that you now have some definitions, some anchors, some numbers. What is their variation in the real world? What are the dynamics of these in the relation to other variables? And what are the relationships between them? So these are really fundamentally unknown things in medicine. So we can measure genes very precisely, but we can't actually measure fens very precisely. And this is really what we're trying to advance. So this is one of our fene assessments, which is the MSRS, which is designed to measure seven domains of MS. And it's a five-point scale. It's been validated against essentially every other scale in the system. And we invented this to sort of meet the patient need. And to give you sort of a quick scan of what we can do with this, this is a set of variability that we can look at. So this is a fraction of reports that are in each level, which is non-mile, moderate, and severe across the spectrum. And you can see sensation, which is largely pain, walking, cognition, upper limb function, vision, speech, and swallowing are all stratified best to worse. We can do some things with this that are kind of cool. We'll look at questions such as how long does it take you to become severely disabled with MS? And this is a finding that came out of this project that's matched to some of the literature, though it's certainly the highest-powered analysis that's ever been done, which shows that you get sicker much faster if you get MS later in life, which suggests there's some relationship to the severity of MS and other changes in your body or context. Now, one of the other things that we study here is how related are these components? So this is a graph that shows that our ability to, using symptoms, guess your severity of MS. So this is symptoms like depression or anxiety. And it turns out that there's actually very, very correlated that with about .82, we can guess the severity of your primary MS symptoms by second-order symptoms. So it's a highly linked complex disorder. Now, that's not uncommon for other illnesses, but just knowing that and that relationship begins to allow us to do some other cool math. This is sort of a standard correlation table, which would say, you know, looking at these sort of 20 fiends that we care about, what is the internal relationship across them. And we're running this up against the data set of roughly 20,000 data points. So these are all significant results in almost every dimension. I'm going to focus for a second on walking, which is just to show that it's not particularly correlated with depression or anxiety, where there are other components of MS that are. And it is highly correlated with dysfunctions in your upper limbs and with stiffness and spasticity. So again, beginning to measure these components that allow us to by factor analysis separate what moves together and what separately. But if you stay focused on walking for a second, and this is where we really move from the description of medicine to math, this is really what does it look like changing? You know, like what, how often do patients stay at one score? How much does it vary? What's the mean? What's the variability? These are all those things that I talked about our theme characterization with regard to walking. And this is probably kind of the richest assessment that's ever been done within a condition of a single sub variable to think about this mathematically. I'm not showing you lots of the sub analysis that was done here. But we even go down to look at individuals. So these are individual patients by numbers and their actual variation in their walking score over time. So you're looking at, you know, highly variant, highly variant disease. And we're trying to understand how to characterize that so that we can hand to GNS and to the other people doing the modeling a characteristic of the health of the patient in a quantified way so that we can match that to genes or change in gene expression or change in protein expression. If I go even further in the walking score here again, this is specifically the percentiles of walking disability over time. So this is a graph that shows at the onset of your disease how severe was your walking impairment and then going out over, you know, roughly three or four years. How much does that progress? And we did this for every component. So you can look for walking or upper limb function or vision or swallowing. And they all progress the way MS patients progress as physicians would describe it. They advance. And you add it all up and you come up with this integrated score, which is what we think represents the total disability of an MS patient across domains. And you're seeing here the progression of MS in the blue lines that shows, you know, where you are and where you're going to go by percentiles. So, you know, we have patients that walk up the 10th percentile or the 30th or the 40th percentile. And you see all the other variables here. So clearly our data matches the clinical description, but we did one other neat thing here that was interesting. We asked what really accounts for this progression? So one of the ideas we wondered was does MS progress the way it's generally thought to progress that there's an injury that increases over time? Or does MS progress more like multi-infarct dementia where you have repeated micro-strokes that actually heal slightly? So we asked if we could map the number of domains injured to the progression. And it turns out that the number of domains actually highly correlated with the progression of the disease, so much so that if you actually adjust the scales, they actually almost perfectly match, which suggests that MS progresses by adding domains of injury as opposed to progressing within the domains that you already have. And when you actually do the math and eliminate the progression, this is the progression within a domain. So it's essentially flat. And if you do a simple mathematical conversion here, which is to say, given that we're measuring roughly seven domains of dysfunction that are roughly representing equal regions of the brain and an equal probability of MS affecting each of those regions, these would have to advance by at least 15% to 18% just mathematically. So this suggests even that the domains are healing and healing quite significantly. So without doing any pathology, just looking at the phenome, it appears by our data that MS is a multi-injury healing model. Now, that may or may not be the case as actually explained at this point, but that's what the data shows from the math. And you can look at this and just looking at domain counts and you can show that more domains is driving the disease across time. And all of this is one of the multiple theories that we're trying to set up for phase two analysis to understand and look at this, both in the way we look at imaging data and the other characterizations that we're interested in. Now, one thing we do in our enterprise is that because we believe in math, you can't describe things, you have to make mathematical predictions. So we build models that say, where will the patient be in the future? Where will they be in one year or five years? What's gonna happen to them? And the nice thing is because we're running a progressive registry, we can check what actually happens in that context. So this is our model evaluation tool and you can see different models against both cheating, which is full regression after the fact. And then the way we can do it without the cheating and see how good our models are as we add more subjects and more data. And this is an internal contest we run all the time. Now what we learned about MS is that it's an incredibly variant disease and we're basically missing some of the fundamental drivers. And so, despite the fact that we're catching all these dimensions really well and we're beginning to understand its variability, we can't really, we don't really understand the primary drivers that are describing the faster variability. But again, specifically gives us mathematical targets and time frames to look at for changes in the disease. But we can do some cool things even though we don't have that. So this is an MS patient. I think this actually is Jess, the one I showed you earlier. This is her background curves against every other patient. So you can see it by percentile. And this is the projection that we can give to her as an individual patient based on the other 5,000 patients we analyzed. And this is our real goal, which is to provide an individual projection for each patient. Now this is not that good. In ALS we actually have a very, very tight progression. We can do very accurate predictions. But it's a start. And if you think about what medicine really is, medicine is the physician saying, you're doing about as we expected, let's stay the course, you're doing better, this is really good, you're doing worse, let's change the treatment modality. This is putting real math on that for the first time. And deployed in the clinic, really what you end up learning is what's missing in your model that drives the understanding of disease. And that's what I think Phenomics is. So I'll wait for questions at the end. Who's up next? I think it's Robert. You're up.