Executive MBA and Sloan Fellows MBA Dean's Circle Reception featuring Professor Andrew W. Lo

good evening thank you for coming I think I know all or almost all of you I'm Dave Schmidt line I'm the Dean of MIT Sloan School it's so great to be able to see this many people not only coming together to hear professor Lowe which is a very good strategy and coming together to be with each other but coming together as you have in support of your school first and foremost I want to say thank you and I hope that you feel that this gathering is a meaningful measure of thanks you're caring about MIT and the Sloan School and for your caring about each other it's important that you've chosen to support the school you know that this is a Deans circle reception and you know that that means that you are collectively among our most important supporters especially giving to the Annual Fund providing unrestricted support to the MIT Sloan School last year I would have been 250 million dollar budget about an 8 million of that money Kait seven million came from unrestricted gifts we have to make a budget that's balanced each year I have to tell you we've been impossible without the support from people like you 80% of the unrestricted budget for the school in terms of giving comes from Dean circle donors it's sometimes the case that you could imagine that there's some whole other group of people that are carrying the loan for the school's need really makes a difference below for the schools need for philanthropic support there isn't some other group what there is is you and it's been great that there's you so thank you very much for that you know that that support lets us help students in need it lets us provide faculty with some of the support that they need to be the thought leaders like Andrew Lowe that make us so proud and they help us invent the future as a school for ourselves new programs new processes for students it's all very important this is also part of a campaign for MIT as you know MIT has set a very aggressive goal not only for financial support but also to make a positive difference in the world as a campaign for a better world I hope you've noticed some of the positive stories and news about MIT there has been a lot of that around sustainability around computing around analytics around entrepreneurship and innovation we're proud of those kinds of elements of impact from MIT not every day is a great day though and I want to acknowledge what many of you have seen today is the day that for MIT the report on the Media Lab and it's accepting gifts from Jeffrey Epstein came out I felt sad and frustrated and embarrassed at the original news about those gifts like many other leaders of MIT we've felt a very strong need commitment for transparency and truth and for significant change both with respect to value and culture and with respect to the processes that are based on those values for accepting gifts I know not everybody's rather a part it's not that I want to turn this evening into a discussion or a description of it but I do think it's that you know that the Institute is serious about the need to do better to do different in its own way to lead other universities because other universities are struggling with this challenge as well how to balance the needs for philanthropy with the need to be able to be the kind of institution that makes you proud so that's probably about enough for me for the moment and so I'm especially delighted to turn to this evenings speaker as such an incredibly esteemed member of the MIT the MIT Sloan faculty as you know Andrew Lowe is the Charles II and Susan Tijeras professor of finance at MIT he's also the faculty director of the laboratory for financial financial engineering if I did even a medium good job of describing the areas not just of finance but of impact in the world in which andrew has been a leader he wouldn't have time to talk and he's a lot more interesting I would just highlight that in settings like using machine learning to help direct financial investments especially in entrepreneurial ventures using tools of Finance to bring capital to the places where it is most needed in life sciences in health care and in the fundamental understanding of the way that markets are in some ways in flux and in some ways efficient through adaptive markets andrew has been in these and many other areas someone who makes me so very proud to be part of the MIT Sloan community so would you please welcome professor Andrew Lowe [Applause] so I want to start by thanking Dave Schmidt life of that very generous introduction and for the privilege and pleasure of being here today with all of you to tell you a little bit about the work that my colleagues and I have been doing you know as an academic I give a number of talks but I have to tell you that the most exciting talks for me are talks with alum because I get incredibly energized by all of the excitement enthusiasm that are here today so I want to thank all of you for coming to this and giving me an opportunity to get back in the classroom as it were I know a number of you of former students and the fact that you're here is even better than those who've not heard me talk and especially given the topic that I'm going to be focusing on it's a rather strange topic and one that finance faculty really don't get involved in and that has to do with biomedicine applying principles of data science to forecasting clinical trial outcomes and I should just say a few words about why I'm going to be talking about this because it's kind of personal and rather unusual and it also goes to the generosity that all of you have shown in supporting the work that we do here at Sloan so first I have to thank Dave for the support of the Sloan School for finance in general and I do that because you see when Dave Schmidt line came to Sloan it was in 2007 and for those of you who are in the financial industry something kind of interesting happened in 2007 right at the beginning of the financial crisis and so when he arrived here in 2000 2008 obviously we were all in the midst of dealing with these things and the Finance Group in particular was dealing with all of the various issues that were happening literally day by day the fall of Bear Stearns and Lehman and the Madoff scandal one thing after another and so you might imagine that a Business School Dean who's starting his tenure would focus on a particular mission to reform the school and think about positioning the school the future and finance would probably not be high on the list of most business school deans in 2007 8 and 9 but Dave actually provided enormous support for the finance group and for me personally in fact oddly enough he asked me to serve as the head of the Finance Group starting around then and it was both the challenge and an honor that I'd never expected and one that took me into some very interesting areas of application the way that he was able to provide that support was through funding various different projects that I guarantee you nobody would have funded I'm not sure anybody would even fund today but you'll be the judge of that after I tell you what it is I've been working on but it's because he was able to direct those funds through the generosity of donors that are in this room that I was able to do what I've been able to do and so I want to thank all of you for making that possible it's a particularly heartfelt kind of a thank you to know that your former students are actually contributing and actually this goes to the whole epstein scandal it's really important to know who your donors are and from my perspective having an alum provide that kind of support means all the world well beyond what the actual dollar amount is so let me now turn to what I've been working on which is applying these principles of financial analysis to health care the personal side of this is that I got interested because a number of friends and family were all dealing with cancer and right around when Dave became Dean and we were in the midst of the financial crisis that was a beginning of a four year period where six people close to me all died of cancer by the end of that I have to tell you I was a bit shell-shocked at one point I felt that somehow knowing me was carcinogenic that I caused cancer and then the more I talked to people the more I realized that you know what we're all in the same boat all of us how many of you here either dealing with cancer yourself or you have a loved one who's been struggling with cancer show of hands yeah it's we're all in the situation and so it got me really frustrated that as a financial economist I was pretty much useless to these cancer patients I had no advice for them I didn't even know what most of the terms meant Charles Harris the donor of my chair called me up one day and said Andrew have some bad news for you I was diagnosed yesterday with stage 4 colon cancer that's metastasized to the liver I didn't even understand what half of those words meant Stage four is that bad it is zero Stage five no Stage five is the morgue and what does metastases mean well it doesn't sound good and I could tell you now having learned the terms that when cancer spreads to other organs that's bad news so I started thinking about all that my friends and family were going through and something really strange emerged from that process it turns out that there's a real contradiction in what's been happening in the field of cancer and the beginning of the contradiction is that biomedicine is at an inflection point and how do I know this I know this because my MIT colleagues in biology tell me so they've told me that we are in a period of convergence this is a report that was published four or five years ago by Phil sharp Tyler Jackson Susan Hockfield and this report is about the convergence of the life sciences the physical sciences and engineering we are now developing enormous insight into how to deal with human disease and the way that they characterized this convergence is what they call the omics revolution genomics the study of the sequence of our DNA epigenomics the on-off switches that cause certain genes to be suppressed and others to be expressed transcriptomics the science of translating the sequences into various different proteins proteomics the study of the twenty to twenty-five thousand proteins that make up the human body metabolomic s-- the chemical reactions that deal with all of these various proteins and most recently microbiome --ax the study of the ecosystem of bacteria that inhabit the human body all of these omics over the last decade have seen tremendous innovation with the exception of one and the omics that has been the bottleneck in getting therapies to patients faster is economics the way that we think about funding biomedical innovation has not changed in 50 years we're still going through the whole venture capital cycle and it's a challenge and so I want to talk about that so I started looking into this and the first thing I did was let's understand how it is that pharma companies make decisions about how to deploy their capital and so because I'm an MIT faculty there's a rule here at MIT that every faculty member in any PowerPoint presentation has to have at least one equation and so so here's my equation all right the cornerstone of what I call healthcare finance is this when you're evaluating a project you calculate the expected net present value because you don't know the real net present value who knows what's going to happen whether a drug gets approved or not next year or five years from now you calculate the expected value of this random variable and the expected value is given by just a few pieces it's the present value of the revenues of an approved drug if you're lucky enough to get that drug approved if you're successful multiplied by the POS the probability of success minus the cost of doing the clinical trials to develop that drug that's it there are not that many moving parts in trying to understand how to evaluate a pharmaceutical project so as an economist I understand most of the terms in this equation I know about revenues I know how to calculate present value and now most of you know too I know I know how to tally up costs but the one thing that I do not know is this probability of success for that I need to talk to the scientists clinicians and engineers and therein lies the problem we don't speak the same language and so in order to make progress here we need to actually bridge a gap and that's the beginning of my work on healthcare finance what I want to tell you about so what is the probability of success well in evaluating these equations I started by looking at the literature in 2010 the Center for the Study of drug development at Tufts University Joe DiMaggio and his colleagues they published a paper saying that across all the various different clinical trials in their database 16% is the probability of success that's a probability that when you start at the very beginning in the laboratory coming up with anti-cancer compounds and you go through the various different phases of clinical trials at the very end of that process which could take ten or fifteen years the likelihood that you'll get a drug approved is about 16% based upon historical averages but then six years later in 2016 another publication was put out this time by bio the Trade Organization for the biotech industry and they claimed that the number is actually more like nine point six percent so in trying to model these expected NPV calculations I was kind of confused is it 16 or is it nine and you know for most people that's kind of close enough but in the pharma industry there is a world of difference between 16 percent probability of success and nine point six I'm just give you an example of how it works I'm gonna give you a specific example of a pharma company called Gilead Gilead is a company based out in California that in 2013 developed an amazing drug sovaldi it's a drug that cures people with hepatitis C 12 weeks of tree one pill a day and boom you are cured permanently of this terrible affliction wonderful drug and later on they got harvoni approved which works even better and in a broader set of patients and so over the course of 2013 14 15 you might expect that they would have made a fair bit of money because of this success so what were their revenues over that three-year period anybody want to guess a billion two billion ten billion ten three years ten billion this is just one couple of drugs one disease ten billion that's wrong try 30 1.5 billion dollars that's what they made in threes it was ten billion in one year yes you're right good good recovery 31 and a half so so now now that you know what these numbers can be what if we actually plugged in the difference between 16% and 9.6 percent what would that mean that's a delta of two billion dollars so based upon small discrepancies in our probability of success it can lead to humongous differences in how you value an asset and in pharma these numbers actually matter they're used to make decisions I learned this in a very personal way when my mother was dealing with non-small-cell lung cancer she had run the the course of treatment that the standard of care really didn't help her and so she was looking into experimental therapies and I was introduced to a very successful biotech company developing a number of cancer therapies one of which might have helped her and so when I met with the CF CEO and the CFO I asked them what I thought was a relatively innocent question does your source of financing have any impact on your agenda does it influence the order in which you prioritize your various different therapeutics and I'll never forget what they told me the CEO turned to the CFO shook his head ironically and said influence our financing drives our scientific agenda it drives it now as an economist I get it you got to pay for stuff but it's the son of a dying patient I was absolutely outraged by this answer what does stock market volatility interest rates and Fed policy have to do with whether you should treat cancer by angiogenesis inhibitors or immunotherapy nothing but those things drive the agenda and it's because of this it's because of numbers like this so it turns out that it matters and so let's try to figure out what the appropriate estimate is so I started digging into how these estimates were constructed using historical data and realize that actually the estimates were being produced by relatively small amount of data but we live in an era of big data so can we do better the original Demasi study in 2010 that was based upon a sample of about 1300 drugs across the 17 year timespan and 50 of the biggest pharma companies that's the data that he used to calculate 16% Michael hey at Al which was the author of the study that was reported by bio they used a much larger sample 5,000 drugs nine years eight hundred and thirty-five pharma companies and about 7,000 clinical trial transition events later on they expanded a data set by Thomas to about a thousand sponsors across nine thousand nine hundred trial events but when I looked at this it just seemed like this was still pretty small potatoes when it comes to really big data so we worked hard to get an academic license from a data vendor called informa and in 2016 we wrote a paper using data for 15 years across 15,000 drugs about 5,700 different sponsors and 175,000 clinical trial events so our data set is one to two orders of magnitude larger than what's out there and what we found is that the numbers are very very different with this larger data set I'll show you in a moment a few examples but I'm gonna give you a website where all of you can get these on your own and from now on on a rolling basis but I want to just point out this is not meant to be a criticism of the other studies why because when Tufts started reporting these numbers and they they started two or three decades ago they've been in the business for a long time when they started doing this work it was big pharma that was developing drugs mostly but that's changed and so I want to illustrate to you what the magnitude of the change is I'm gonna show you the list of the top 30 drugs in terms of revenues worldwide back in 2000 so you won't be able to even see this list from where you are but the key thing that I want you to focus on is the rows that are highlighted in blue because those are the drugs that came out of biotech companies smaller companies that works crappy startups and academic medicine so out of the top 30 best selling drugs worldwide in the year 2000 exactly four came out of biotech and academia and if you look at the top five none of the top five came out of biotech in academia what that means is that the source of these drugs is Big Pharma and that's why when Tufts did their initial analysis and the analysis that they currently do it's with the 50 largest drug companies fast forward to 2015 2015 can anybody tell me how many of the top 30 do you think came out of biotech or academic medicine 12 pretty good guess that's three times more than in 2000 try 2020 and out of the top 5 how many big pharma companies produce the top 5 best selling drugs 0 biotech and academic medicine now this is not meant to be a criticism of Big Pharma because the reason for this is not that anybody failed but rather there are a lot of people that are succeeding who are succeeding academics scientists clinicians engineers the this is the biomedical inflection point that I was talking about the omics revolution and it has changed the economics of the business so in order to produce better forecasts we decided to use a larger data set and we also dealt with a number of data issues that those of you who are data scientist and I already met two or three of them during the cocktail hour data scientists will know well which is missing data it's very very messy data and you have to deal with all sorts of very subtle issues and having done so we actually published a paper just last year in a journal called biostatistics on exactly this issue how to estimate probabilities of success now I have to stop for a minute and just explain a little bit about this situation because it's a really weird thing to be able to publish a paper on how to calculate probabilities of success particularly in a journal called biostatistics up until recently I had no idea what this journal was but people tell me this is a very respectable journal to a top-tier journal for this field why would they publish an article about calculating probability of success where is the theoretical underpinnings the the deep statistical analysis here we do use a very important mathematical tool in order to calculate our probabilities of success it's called division but it turns out that missing data and how you account for the various different clinical trial phase transitions is not trivial and so they published the paper and more importantly we negotiated with the data vendor to be able to make available our estimates for free to the public so we launched project alpha last year since I teach at a management school I've got to come up with snappy sounding acronyms and project alpha stands for analytics for Life Sciences professionals and health care advocates and what it is is a website which I'm going to just go to now very quickly it's project alpha that mit.edu you can all log in to it and this website allows us to basically report probability of success estimates on a rolling basis using the data set that informa has provided so for example our last estimate was third quarter of 2019 we just got the data for fourth quarter of 2019 so within the next three to four weeks you're gonna see a refresh of this we've got probability of success by various different therapeutic areas we've got that phase by phase we've got estimates of rare diseases we've got estimates with and without biomarkers and we've got estimates time-series estimates that can show you over the course of the last several years are the probabilities going up or down in fact very happily because of the omics revolution they've been going up so all of this now is available at any of your fingertips if you want to plug in the POS number you can actually do that from the source but we can do better and let me explain how so in addition to calculating historical probabilities of success using division we can actually set ourselves a more ambitious goal of figuring out how to alter the probabilities so it turns out that one of the probabilities that you didn't have time to look at but or leisure I would encourage you to do so what's the probability of getting a cancer drug approved historically it turns out about 4% terrible odds it's a little less than one out of 20 so there are 20 balls in this urn the red ball is the actual anti-cancer compound that's out there and you got to put your hand in this urn and draw that red ball 19 out of 20 times you will fail and that's not good news for cancer patients what if we could alter these odds in particular what if we could maybe find an extra red ball but more importantly get rid of some of the yellow balls what if we could actually tilt the odds by reducing some of the duds what if we could forecast better the likelihood of clinical trial success well it turns out that with modern AI we can do that and in a paper that we published just about six months ago we actually use machine learning algorithms along with statistical imputation to deal with missing data to be able to predict clinical trial outcomes so not only using backward looking data to calculate the odds historically but using all of the features of a clinical trial and a drug indication pair can we actually handicap the outcome better this paper has been published in the Harvard data science review new journal focused on data science why Harvard well because MIT can't do everything so I would encourage you to take a look at this the method that we use is pretty standard kind of machine learning random forests support vector machines decision trees we looked at a bunch of them and if you're interested in the paper we actually spell out everything that we do not only do we spell out everything we do we actually provide the software with which we do it so you're welcome to take it use it it's got an open-source license so you can all access it what we're doing is essentially trying to run a forecast and the forecast makes use of features different characteristics of both the clinical trial as well as the drug itself the drug features that we're using are the route of administration is it a pill is it IV the particular medium is it suspended in a liquid and a gel the biological target target family are you trying to target a particular protein and enzyme some kind of of RNA AI and so on in terms of the clinical trial features that we're using to forecast these outcomes the duration of the trial the study design is it a fixed sample clinical trial is an adaptive trial and what kind of a sponsor track record do you have the trial outcome and so on and so when you put all of these things together and you forecast and ask the question are you doing a good job you won't understand how to interpret these numbers until you start delving into the whole machine learning literature but a you see this measure of area under the curve measures between 0.5 which is random no good at all in terms of predictions vs. 1 which is perfect predictor and you're seeing a ucs of somewhere between 60 to 80% what this is telling you is that by no means you get a perfect prediction but it's not random either it actually provides you with incremental information and here's an example so what I've got here is a histogram of actual forecasts from the machine learning algorithm the higher the forecast the more you are to the right the more likely it is that you're predicting approval okay and the color coding is what happened after the predictions were made so we did this as of 2015 and we're looking at what has happened in 16 17 18 green means the drug got approved red means the drug failed and white means we don't yet know and what you're seeing is that the higher the scores the more green the lower the scores the more red it's not perfect by any means but it's showing that actually these machine learning algorithms are having some kind of predictive power and the important variables that we found in the prediction are listed here trial outcome trial status prior approval for another indication so on and so forth I won't take you through all of these but I'm gonna give you one example of a trial characteristic that surprised us because it actually mattered in ways that we didn't anticipate it turns out that for a clinical trial the speed of recruitment actually matters for predicting the success or failure of it if you recruit a clinical trial faster meaning when you open up a clinical trial you invite patients to submit an application to see whether they're acceptable for the trial the faster the patients come into the trial the more likely it is it'll succeed why is that why should it have anything to do with speed of recruitment well we found out after we talked to a few doctors it turns out that when an experimental drug is given to a patient their doctor knows about it and has to oversee it if an experimental drug works really well you will know in the first few days or weeks that it's working in many cases and if it's working your doctor finds out about it and if your doctor finds out about it he or she tells all of their doctor friends and now all of a sudden instead of one patient that wants to recruit you've got 20 to 30 that are knocking on the door of this clinical trial so that's an example of a simple statistic that we didn't know mattered but anecdotally doctors knew clinicians knew and so when you put all this together you're actually making these predictions and so for any of you who are data scientists if you want our code it's available on github you can actually download it we also have a link at project alpha and our feeling was put the code out there people will start using it and this will help patients not quite that simple we discovered that many pharma companies do have data scientists but they're busy focusing on applying these techniques to actual molecular design the clinical trial aspects generally are done by biostatisticians and that group is not yet as familiar with these methods but we got some very interesting comments we got requests from pharma companies that said we would like you to do this for us if you don't mind and my response was of course be happy to do that send us your data and I'll have some students working on it and we'll publish the paper and the response was no no no no no no no that's not how it works we're not sending you anything we want you to come here and use our data as well as yours and you're not publishing anything we want to keep this proprietary the problem is we can't do that our students need to graduate they need to publish their theses and as an academic we need to use the precious funds that you've donated to us in a way that will be lit will allow us to disseminate the research so this is where practical applications come into play one of my postdocs and PhD student show mesh Chaudhary a number of you know him because he was the TA for the very first healthcare finance course that we taught here show said look we got to do this because if we can show them how to use these tools that's gonna actually you know help patients in the long run so show decided that he was done with the postdoc and wanted to actually apply these in practice so he started a company with me called quantitative life sciences advisors and we actually now produce this on a commercial basis so I'll give you a very quick demo of the site because we're really excited about it we're hoping this is going to have impact in the industry so this is a depiction of clinical trial data that I won't take you through because we're short on time but I want to show you exactly how our machine learning forecast work so in this table that most of you cannot see it's a list of all the clinical trials paired with specific drug and pharma company sponsors that are currently in process so this is not historical data this is all live and what we've produced our machine learning forecasts of the probability of success in this column to go to the next phase and the probability of going all the way to approval in this column and you can sort this list by the kind of drug that you're looking at by the disease by the therapeutic area even by the pharma company and if you want to see what the graphical display of these probabilities look like it's given here so each of these bubbles is a separate and independent clinical trial and it's arrayed horizontally so that the probability of success of zero is to the very left and the probability of success of 100% to the right so the farther right you are the more likely it is that you're estimating success the color-coding indicates the phase of clinical trials phase one is the very beginning where you're just looking for safety Phase two is in blue where you're looking for efficacy in a smaller patient population and Phase three is the big study where you're looking at whether or not it works in a larger patient population and so let me just show you what we're doing here if you hover on this green that's a phase three trial for an asthma disease 802 patients so the size of the bubble gives you the number of patients in the trial on a logarithmic scale and you'll notice it's a sponsored by Novartis and you'll notice that there are a list of features in green those features are telling you what the machine learning forecast views as the most important factors and green means the most important positive factors in producing the forecast and if it's red it'll tell you that they're negative factors in predicting that you're not going to succeed so we've done this for every drug indication parent that's on the market and you'll notice something interesting when you go down here that crowded area is oncology there are a huge number of cancer trials out there and the thing that I really like about this platform is from the financial perspective it allows you to start thinking more systematically like a financial investor let me show you why so I'm going to pick a pharma company like GlaxoSmithKline because they are big and they're across lots of different areas so this graph is GSK s entire portfolio so you can see in one picture all of the things that they're doing and the risks and this is exactly the point of financial engineering it's to stop looking at these one at a time and just start putting it together in a way that is systemic so the last thing that I'm going to tell you about is what do you do with these probabilities of success well you make use of it in that equation that I put up so as part of the the service that that show has developed it's to create a spreadsheet that allows pharma companies and biotech companies to pull in all of these different estimates and apply them to NPV calculations so here's a spreadsheet that calculates the total revenue approved drugs NPV of all assets for two thousand twenty twenty one twenty two so this is a pro forma spreadsheet that a biotech or pharma company would typically have in looking at their portfolio and you can see here in this spreadsheet the pipeline of projects I've listed 14 here this is just a sample and green means the drug is approved red means it's been it's failed and any color in between means it's in process and the probabilities that I showed you from machine learning exercises is pulled right in here now the best part about the spreadsheet is you can actually run a simulation so let me push this button and you'll see what's going on we're running two thousand different draws of those urns and these drawers are basically showing you over time what's happening to your portfolio as these clinical trial events read out so if I go back to the pipeline you can see that the various different draws are happening so now you only see either red or green either it succeeded or it failed but Oh with 2000 different scenarios it's running these simulations and not surprisingly your cash flows are very depending on those drawers and as the simulation complete so we just ran in real-time this is actually the spreadsheet that's running in real time you're actually seeing 2,000 of these simulations occurring after these simulations have completed you get a summary and the summary shows you the expected NPV of the portfolio but this graph which is one of my all-time favorites it's showing you something more than that it's showing you not just the expected NPV but the confidence band between the 5th percentile and the 95th percentile it's showing you the risk how much are you betting on various different milestones and if you go all the way to the end you'll see something really cool which is the histogram of your cash flows it turns out that this histogram is trimodal and it's telling you've got some lumpiness in your portfolio and you need to manage that risk carefully so show built all of this in a matter of months we started the startup just in June and imagine if you now include all of the data that pharma companies have including what we have from informa that's really powerful we're actually doing that one of the one of the clients that we have is a pharma company and very interestingly they said you know they're the ones that said you know no we're not gonna send you any data in fact their data are not on the cloud they've got an air gap computer where all of their data reside precisely because they don't want to risk getting hacked and so of course what we what did we do we built a server that is an air gap computer that they can put their data on and be able to calculate these prices in a fashion that it's totally secure so let me wrap up by pointing out that for the last forty five years we have had a war on cancer and I don't know whether we're winning or losing that war I'm not qualified to say but as an economist I can tell you that war is the wrong metaphor because war is based on hate fear and you cannot live in a state of hate and fear for very long those are not sustainable emotions greed on the other hand is very sustainable and so if we can use finance to further the goals of biomedicine we can actually have a huge impact finance doesn't have to be a zero-sum game if we don't let it we can all do well by doing good and thanks to all of you and your generosity we can do it now thank you and stuff time for a couple of questions yes Oh actually we won't wait for the microphone oh no just if you don't yeah so I'm curious about novel molecular targets how do you weigh those when for example a pharmaceutical company is looking for a cancer drug and they are very focused on a molecular target that has not been looked at before how wait that so we don't wait it but the machine learning algorithm implicitly waits it because one of the features that we're using is whether or not this is a new molecular or new biological entity and there's no doubt that when it is new the probability of success is lower so one of the things that we did was to ask the question you know repurposing drugs is that a good thing or a bad thing all of you who are in the industry you already know the answer it's pretty clearly a good thing other than the fact that there's no way to make money on it if the drug is generic but in terms of probability of success absolutely a new molecular entity a new biological entity is actually very difficult to get approved now having said that one qualification is that it turns out that if you've got a drug that's approved already for a given indication and you're predicting an update and an additional approval for a related indication it turns out that that actually goes both ways because if it's already approved then the hurdle for getting approval for the next indication has gone up because the standard of care has just gone up so you'd not only have to beat you know a placebo you actually have to beat the next best thing but but overall you're absolutely right new molecular and biological entities have a negative impact on forecasts of probability of success maybe one last question yes here fascinating presentation thank you very much I work for a large former I've worked for large former for about 20 years and I think as much as I have learned their process or the the the work that goes into developing a new drug has changed a little bit my company we have I think north of ninety thousand employees and so we recognize that actually the best ideas are not within the company right and not all the worst ideas within the company either but what that means is that there's much more collaboration and ever was both with academic institutions and with smaller companies and so I just wanted to comment on that and just understand a little bit better your point at the very beginning of delineating small large academic or not doesn't is that's not really the point in my view I think the point is we need to work together where the strengths are right right and a lot of discovery has long been recognized to be not in a large pharma however it does take all players to bring it would you would you agree I agree a thousand percent with that comment that's incredibly important and thank you for bringing that up so the purpose of showing you that comparison of 2000 and 2015 was actually to make that point but let me talk about the financial implications of what you just said in 2010 Morgan Stanley put out a rather controversial report about the pharma industry and the title of the report I still remember it because it's seared into my mind the title of the report about the pharma industry is this exit research and create value the point of the report was that big pharma instead of doing more R&D they ought to do less Rd and more M&A and the reason is not because they're necessarily bad but it's because the complexity of biomedicine has grown dramatically over the last 10 20 years and so no one company has all of the intelligence and experience to be able to develop drugs in fact we don't need generally know where the break the grain ideas are gonna come from the very first designer drug Gleevec imatinib was approved in 2003 does anybody know where it came one came from who the main academic was who developed that drug you think it was somebody from Harvard MIT Stanford those would be good guesses right the person was Brian Drucker who was a faculty member at the Oregon Health Sciences University now I I hadn't heard of that University until I read about Brian Drucker but that goes to show you the complexity of the field and so to your point it has to be a collaboration between Big Pharma little Pharma biotech academia how are we gonna get all of these disparate parties to collaborate that's financial engineering that's what we are hoping to do over the course of the next years with coming up with new methods of financing this kind of process so on that note I I don't want to run over thank you very much for your patience and for coming here it's been a pleasure and a privilege and I look forward to staying in touch with all of you thank you [Applause] so I hope you all found that to be as inspirational and enjoyable as I did could we thank you Andrew one more time please for being here with you I know we're very short on time and there are I hope still beverages let me just say thanks for gathering thanks for being here please keep supporting the school if you would like to support work like Andrews specifically we have a variety of matching challenges see any of the staff here but please care about each other and care about MIT and have a great evening thank you [Applause]

Channel: MIT Sloan Alumni

Views: 916

Rating: 4.8400002 out of 5

Keywords: #MITSloanAlumni, #Faculty Reseach, #Andrew Lo, #Financial Engineering

Id: gaRnpJ2JepM

Length: 51min 6sec (3066 seconds)

Published: Tue Jan 14 2020