a16z Podcast - a16z Podcast: The Science and Business of Innovative Medicines

Episode Date: January 14, 2019

with Vas Narasimhan (@vasnarasimhan), Jorge Conde (@jorgecondebio), Vijay Pande (@vijaypande), and Sonal Chokshi (@smc90)On average, only 1 out of 20 medicines works when we actually bring them into the human body, and these rates of success haven't moved much in the pharma industry overall in the past 15 years, despite much scientific progress. Because if you really think about it, it's incredible that we find any human medicine that works at all, given that human beings are the product of billions of years of evolution, and represent an incredibly complex system we do not fully understand. Yet the business of the pharma industry -- and Novartis in particular, which covers everything from generics to innovative medicines -- is not that different from other large enterprises when it comes to managing R&D and pipelines of ideas, talent, and sales.So in this conversation, a16z bio general partners Jorge Conde and Vijay Pande with Sonal Chokshi interview Vas Narasimhan, CEO of Novartis. How does the world's largest producer of medicines in terms of volume -- 70. billion. doses. a. year. -- balance the science and the business of innovation? How does an enterprise at such vast scale make decisions about what to build vs. buy, especially given the fast pace of science today? How does it balance attitudes between "not invented here" and "not invented yet"?Narasimhan also takes us through the latest trends in therapeutics, such as cell and gene therapies (like CAR-T for cancer and more); RNA-based modalities; and others -- a sweeping tour from small molecules to large molecules to proteins and other modalities for making medicines. But where does tech come into all this, and where are we, really, on science becoming engineering? Why do both big companies and bio startups now need to get market value signals (not just approvals!) from payers earlier in the process of making therapeutics? And beyond all that, how could clinical trials be reinvented? Finally, what should all scientific (and all technical) leaders know when it comes to leadership? All this and more in this episode of the a16z Podcast, recorded recently on the road while at the J.P.M. health conference in San Francisco. image: Global Panorama/ Flickr

Hi everyone, welcome to the A6 and Z podcast. I'm Sonal. Today we have one of our special episodes live from the road. Well, not really live, but recorded here during the JPM conference in San Francisco this week. So in this conversation, A6NC Biogeneral Partners Jorge Kande and VJ Pande, join me to interview special guest Voss Neurosimhin, the CEO of Novartis, which is one of the largest healthcare and pharmaceutical companies in the world. In terms of volume, they're the largest producer of medicines with 70 billion doses a year across a wide range of therapeutic areas from cancer to cardiovascular disease and more. So in this episode, we cover everything from the latest trends in therapeutics,

from cell and gene therapies to RNA and proteins to other emerging areas and what's science versus science fiction. We also briefly touch on clinical trials, go to market, talent, hubs, startups working with big pharma, and throughout we consider where tech comes into all of this and what happens when science becomes engineering. But we begin with the business of science and innovation, both inside and outside. You build up R&D expertise in our industry over long periods of time. If you think about cardiovascular disease, we've been in it 40, 50 years. And you think about transplant and immunology, again, 40, 50 years, oncology, 25 years.

So you build up an accumulated expertise. And really the art of it is to make sure you have a depth of new medicines to keep filling your pipeline in each one of those therapeutic areas. Now, there are instances where we find new breakthroughs and areas we're not in. Those you have to really think about, are you going to really stay in that area for the long term? The other element of the story is when you really have exhausted your pipeline. We're not so good as an industry at this, but you have to also be prepared to exit, I think, areas where you're going to be subscale. And that's something we're working on. We've made a number of exits actually this year where we just said this is areas we just can't sustain longer term.

Can you give us a little bit more color on how you make those decisions, especially as a CEO steering this? that's a pretty big, I mean, frankly, it's a decision that every big company, regardless of industry, has to think about, which is essentially what to proactively invest in and what to proactively opt out of, which killing things, as we call it in the media business is a pretty hard thing to do. How do you think about that and how do you tease apart the signal from the noise when you get a lot of inputs, both internally and externally? So we do it currently at two levels. One is from an overall portfolio standpoint. We've made the decision to really focus as a medicines company powered by, advanced therapy platforms and data science. So in order to really make that happen, we transacted in 2018 around $50 billion of deals to really change the shape of the company. We took principal decisions to leave consumer health care because we just didn't believe we would be a long-term leader in consumer health care, a decision to spin our Alcon business, which is to get out of medical devices and contact lenses. And alongside that, as we moved out of those other areas,

we made significant investments of acquisitions in this next wave of therapy, cell therapy, gene therapy, in an area called radio drug conjugates, which is a nuclear medicine kind of area. So that was at one level, at the portfolio level, changing a 20-plus-year trajectory to actually become a very diversified company. It really came out of a conviction in my mind that science is moving so fast. You have to focus your capital and really focus your energies. that's at the macro level. Now, when you zoom into innovative medicines and we have to decide, okay, which therapeutic area, do we stay in cardiovascular disease or do we stay in ophthalmology? I mean, those are pretty tough decisions because if you take down an R&D effort, so one example

for us was infectious diseases where we had a longstanding effort. It was not an easy decision. I mean, it was a lot of going around. Are we really sure? Because you can't change your mind now in three or four years and say, I wish I had it back. It'll take you another 10 years to build it back again. The cycles of innovation and science are accelerating. The science is moving much more quickly and it has in the past. So assuming that that's the case, will it continue to be true that it will take decades to build up expertise in any given therapeutic area? In other words, will there be future emerging players that come much more quickly than they have historically? I think there can be very fast players who are really working on a couple of medicines or a couple of assets. But when I

talk about building up a capability. I'm really talking about a scaled capability that could generate new medicines consistently over time. And while I do believe the pace of science is improving dramatically, we also have to keep reminding ourselves and being humble with the fact that we understand a fraction of human biology. And actually, when you look at attrition rates in our industry, really the chances of success that we have, they haven't moved in the last 15 years. Still, when we bring a medicine into human beings, on average, only one. out of 20 works. Really?

And that has stayed constant, despite the fact we've had this explosion in new science. I'm going to quickly pause on that for a moment because that's a pretty important point. So only over the last, say, 10, 20 years, only one out of 20 medicines actually work in the human body. Once we get it into human beings, we have about a 5% success rate, 5% to 10% success rate. And it varies by therapeutic area. We've actually been fortunate at our company.

We average in that same metric about 8% to 10%. But if you look industry wide. it is about 5%. The nutrition rates are pretty constant, but the costs they'll keep blowing up, too. They do. How does that work out? I mean, because in a sense,

you know, one analogy that people use is almost like trying to get oil out of the ground. And, you know, the low-lying fruit, I'm mixing analogies here, but the low-lying fruit has been taken and it's just harder and harder to find new therapeutics? Or do you feel like the science is moving fast enough

that that's not an issue? You know, I think we go through waves. I think there was a period of time where probably in the 1990s and early 2000s, we had a pretty big wave of innovation and we'd bring a lot of medicines forward. We went through a lull for seven, eight years. Now, I think with, again, explosion and ability to really understand the mechanisms of disease, we're seeing a renaissance, a record number of FDA approvals.

We're investing heavily in new therapy areas. I mean, 15 years ago, people would have said you're crazy if you think we're going to do gene therapy and cell therapy and all the things now that we're doing at scale. The costs really come from our ability to manage complexity. When you look at it, over time, the trials get more complex. The requirements from regulators get more complex. Because the science gets more complex, we can actually measure more things. So we add and add and add and add.

And that's led to an interesting, pretty linear increase in cost per patient in our clinical trials. I don't think it has to be that way. I think what really our industry has not been great at is really deploying technology to make this much more efficient. So I think there's a lot of opportunity. Why do you think that's the case that's been so hard to deploy? I think it's, you know, we're a high margin industry. So unless it's easy enough just to keep arguing to yourself, it doesn't really matter. As long as we get another big medicine out, it's okay.

Let's just keep going. Well, and if you screw things up, there's a huge cost. There's a big downside. But I think now we're reaching the point where we have no choice but to really now engage technology. I mean, there are estimates now from various sources that believe you could take out 20% of clinical trial as costs if you were actually to really deploy technology at scale. If the attrition rates have been flat for as long as they've been, and there have been all of these proliferation of new platforms, self-therapies, gene therapies, is there a measure

that you qualitatively or even quantitatively can look at that says that the medicines that are getting through are meaningfully better medicines or different medicines? In other words, the failure rate might be the same, but the impact of success is it greater now in any measurable way? So it's a very important point, and there's no objective measure. I mean, various institutes have different, different measures, but it's nothing I think that is used externally. We internally have just set a very clear bar now for ourselves, primarily because we live in a world where now where nobody wants a Me Too medicine or a medicine that's just

incrementally better. We say to ourselves, it has to replace the standard of care. And that usually means it gives such a big clinical benefit to patients that it It just becomes the de facto medicine of choice in that therapeutic area. That's a shift. It means a lot of projects no longer make the cut because you're really asking yourself, if I don't have something really transformative, I'm not going to take it forward anymore. And so all of our research teams and development teams are having to now come to grips with that,

that we will stop projects unless we really believe it can redefine the standard of care. I have a process question behind this because it's parallel to this idea of basically going for slugging average versus batting average and like outsized hits with great outsized impact. So behind the scenes, what are some of the mindsets that you and the R&D teams bring to bear to make these investments for slugging versus batting average? How do you set things up to make that happen? We have all of the various review committees and portfolio meetings, et cetera. But really what it takes is a lot of discipline about the criteria that you're using.

So we have very clear criteria. We try to apply that rigor. I think people, there's a lot of romanticization and R&D about big ideas. So much of it is just about discipline and discipline. The nitty-gritty. The nitty-gritty, disciplined execution of how you look at projects. I think that's one element. I think second, you have to build patience because part of the reason mediocre projects go forward

is you start to worry you don't have enough in the pipeline. And you start to lose faith that something's going to come. And you have to believe in your own scientists and your own R&D engine to say, I'm going to say no five times because I believe that. the sixth one could be the big one, rather than get worried and just start letting things through. Because actually what you do is you crowd out the money. It's an huge opportunity cost.

It's a huge opportunity cost when you take those. And that's been a real ongoing challenge for us. I think that the third element is to bring a real lens of what does it take to be successful in the market. I think historically we just had a belief that if we had a great product, it'll all work itself out. Now we actually ask the market access teams that have to negotiate with payers

to show up at every meeting and say, actually, even in phase two, so really early for us, what is it really going to take to, let's say, bring a new medicine forward in asthma, a new medicine forward in multiple sclerosis. And if we don't make the cut, we just have to be brutally honest with ourselves. My reimbursement's more important, or more on your mind than sort of just getting past FDA. It used to be we think about reimbursement as we got to launch. Now we're thinking about it really early in development. For people that are new to the space or just a lot of entrepreneurs,

think that the FDA is the real challenge, and just getting something clinical trials is expensive and hard, and that's true. But the reimbursement, being first in class, being having this huge jump in care, that is the real challenge. And so what I would love to see, especially in our founders, is for them to work backwards. But work backwards, not from getting through trials, but work backwards from reimbursement. Yeah. And the way that Voss describes it, I think, is absolutely true is a lot of people view reimbursement as a process to get to market access. But reimbursement is really just a proxy for value proposition. So what are the actual user stories? Who's going to actually value this? Who's willing to pay? It's almost like a pricing

study. It's almost like a price discovery in the consumer world. In this case, it's, you know, obviously the payer is not the direct beneficiary of the therapeutic, but they do bear the burden of the cost. And so they're the great arbiter of saying, is there a true value proposition? And actually that's why when you talk about moving away, industry moving away from Me Too drugs, it was because a me too drug arguably could not show a very significant marginal increase in value proposition and therefore you could be very difficult to justify an increased premium price. And so that historically has been the big challenge. On that note, I do find it ironic.

A big part of your business is still generics. So, I mean, what is that but a me too drug? Like, how does that fit into this big picture? Yeah, so, you know, if you look overall at Novartis generics, you know, from a sales standpoint and a value standpoint, is a small portion of the company. But you look at a volume standpoint, it's the biggest part of access. And so really what our generics business does is take when medicines go off patent, we then produce them at scale.

We're the largest producer, for example, of penicillins in the world. So we have a huge role to play in providing access to medicines around the world. I mean, right now, Novartis reaches about a billion patients a year through our work. And a lot of that is through our Sandoz generics unit. So if you break it down, so if you, there's 70 billion doses, that are Novartis drugs every year, how many of those 70 billion are generics? I would say roughly 80%.

Is it standard that big pharmaceutical companies have their own manufacturing facilities? And do you see that changing anytime in the near future? Most pharmaceutical companies have their own manufacturing. I mean, there's different trends right now. There's a pretty significant increase of use of Chinese and other producers from many elements of the manufacturing, but still historically we've had our manufacturing facility.

the biggest trend we have right now is a shift to these advanced therapy platforms so what we're having to do is as our volumes go down and kind of the old older medicines that were produced in huge volumes and innovative medicines we're now building up cell and gene therapy production facilities around the world so that's a that's a shift we're seeing you know you talk about novartis becoming a medicines company using data science and novel platforms you're very specific about saying medicines are medicines and medicines and therapeutic synonyms in the Novartis mindset? I would say yes. There's, of course, a gray zone here. So, you know, what is a therapeutic? We would say, you know, medicines is our proxy for therapeutics. I mean, one example, we launched in the U.S. a digital medicine.

I mean, with paratheraputics, this is the first digital app with an FDA label that's being used for opioid addiction and other psychiatric illnesses. And it is literally an app that has run clinical trials. and has gotten an FDA-approved label. So that's truly an example of a therapeutic. But I would put that within our world of medicines. So software is a drug. Yeah, software as a drug.

Most surprising indication that you would expect to see for a digital therapeutic. Because I think most people assume that it's going to be around, you know, behavioral health issues or addiction, like with the work you've done with Pair. Can you imagine moving beyond that from an indication standpoint for digital therapeutic? I mean, my hope would be we could develop one for obesity, right?

That somehow that a digital therapeutic that could actually just move the needle a little bit more on obesity. It's such a massive issue for society. And it should be one where a behavioral intervention on top of other interventions could actually move the needle. Because so much of it is behavioral. I mean, there's not an example that's non-behavioral in your future. You're not curing sickle cell with an animal. I mean, I wouldn't put a guess around fertility, but one could argue that's also psychosomatic. Well, I mean, the thing is that actually, so you think about like the modern medical sort of marvels, I think about like an antibiotic.

Like I was sick when I was in college and I had a super high fever. I got an antibiotic and like next few days I'm fine. Maybe without that I've been dead. And so that's kind of magical. And it's not like I have to take antibiotics for rest of my life or whatever like that. I'm just cured. But, you know, the amazing thing about behavioral is that that's where you don't have this. I can't imagine that you have a molecule that cures depression.

You take that and then you're just done. Or you take a couple doses and then you're no longer up type 2 diabetes. And behavioral is really broad. It's depression. It's smoking sensation. It's type 2 diabetes. It's even quite possibly Alzheimer's. I don't know if you've seen like all these.

I've seen a lot of recent papers on this. It's fascinating. And so these are actually the areas where if you look at the biology of Alzheimer's disease, that's just a mess. You know, so it could be that for these things where you have a very clear target. I just have to hit the ribosome. the bacteria and then we're done. That's easy. But there may be actually the future of things where just it's hard to hit with a molecule. And all that is primarily behavioral. Interesting. So basically, you're almost arguing the question might be moot because all of disease

is behavioral in some capacity. Well, no, all the stuff that's hard. The complex is complex. The low-lying fruit molecularly is not behavioral. There's this infrastructure layer that's being created now around gene therapies. So as folks figure out manufacturing, as people think about delivery, as people think about, you know, all of the various components of modular aspects. Do you think those are things that necessarily would be owned by one company or these horizontal infrastructure layers that, you know, a third party should develop and sort of deploy across the industry?

How do you think this plays out? In other words, is there a startup that figures out AAVs? Do they sort of supply AAV to the industry or do they go and develop their own gene therapy? It's a very timely question. We don't know the answer yet. I think right now in this nascent phase that we're in, we believe we need to just own it because the launches are so important that we can't afford there to be a lot of experimentation and not really owning the supply chain.

We've done $15 billion of acquisitions just last year in the space, not including all of our internal work in each of these areas. So we've chosen to build out the infrastructure ourselves. I think as the technology matures, we'll get more comfortable about, which areas we could send out. I also think the entrepreneurial world will also figure out where they can play a role. I think that's still all being figured out right now. And I actually don't have a view yet. I don't know what's going to be the elements we must own and what are the elements that we could afford to give to other parties. You know, on that note, I'd love to

hear from you more about how you figured out the build versus buy piece then, because a big part of your work is, you know, focus on innovative medicines. And you made this argument that it takes 10 years to build up a base even longer, 20, 30 years, and yet you're also acquiring the expertise for the very new cutting edge things, which almost makes it seems like you don't seem like you don't have to even bother building up that base.

Why not just acquire it? So how do you sort of navigate the build versus buy part of this? I think when you want to enter very new areas, sometimes it's prudent to ask yourself to somebody, somebody have this much more figured out than you do. So if you take the example of gene therapy, we acquired a company called a Vexus, really, I think, the front leading-edge gene therapy company.

Now, the scientists at AVEXIS, you know, they've been working at this actually in their academic labs for 25 years. I mean, they've been working on trying to hone how to use AavV vectors to get to the neuromuscular system of children to address these issues. They'd actually figured out the manufacturing. They'd built the manufacturing site. We were working on gene therapies ourselves in-house, but when we looked at that, we said, this is an opportunity to really accelerate what we're doing.

And so it made sense, I think, to go external. There's always that balance. You know, we are a company that's very focused internally on research. We consistently invest at the high end on internal R&D simply because we believe that's the heart of the company. But what I'm trying to keep asking our people is if there's somebody out there who's got it better than us, let's just go get that and then we'll build off of it. I love that. But there is a classic NIH non-invented here. syndrome. And when you have a strong internal R&D culture, it does compete with NIH a lot.

So the question it really begs is how you then, with all these amazing acquisitions, integrate them into the company and actually make sure the classic Chesbrough study of all these acquisitions not being killed by the big company, like how do you balance that piece? So I think there's two things I'd say. One is as a R&D person, I have sort of the ability to really get in there and have the discussions directly with the scientists and argue. why we need to actually go external and really evaluate the case with hopefully objective eyes. The other thing we've decided to do, at least with these very new tech, three new technology platforms, is leave them as independent units and really let them grow up independent from the big R&D and manufacturing machine.

Because I think exactly for that concern, it makes sense to let them build up and really, you know, incubate these new technologies, get them all sorted out. And then we can ask the question, what's the right setup down the line? Right, right, right. That is what the classic studies show. That sort of is the way to do the success. I did, by the way, find it very fascinating because I wasn't aware that you have a scientific background. It reminds me of this idea that we have around CTO-led, you know, really having technical people at the helm. So I am curious about your view, I mean, besides being able to talk to the internal scientists, like, how has that affected your own career and trajectory at Novartis so far? Given the company's heart is innovative medicine, and most of my background has been in drug development and really developing vaccines and then developing various medicines. I think it gives me a really good insight into the heart of the company, our key technology. If you think about our pipeline today, I know every asset, every clinical trial,

I know all the clinical trial endpoints. So that, I think, gives you a certain insight into where the company is heading. And also, I think, enables you to hopefully guide the company into the right area. in the future. I think it would be self-serving to say that it's better to have an MD, R&D person running companies. But I think it does give you a different perspective on an R&D industry like ours. Right. It might even be able to help to be able to empathize when you are killing a project that you actually know what it's like to feel that.

Well, that's for sure. Okay. So on that note, what are some of the most interesting and most innovative medicines categories? You know, when you look broadly right now, I think you're seeing, you know, a few big areas of high innovation. I mean, I think in the whole world of CART, so cell-based therapies, really what this is is harnessing the power to take cells out of the human body, reprogram those cells and put them back in the human body. CART is the way we do that in cancer, but there's certainly the opportunity to do that in many

other diseases. There's companies working on trying to cure sickle cell disease, others working on other inherited disorders. So really reprogramming cells. So if you go back 10 years ago, something like a CAR-T therapy would have seen science fictiony, or at least maybe 20 years ago. If we look forward 10 to 20 years, what are the modalities of the future, do you think? I think a couple of things will likely come. I think xenotransplantation, I mean, which has been in and out and worked on. And what's interesting is every one of these comes up and down. So, you know, gene therapy, cell therapies popped up in the 90s, kind of went away, popped up in the 2000s.

kind of went away. And then the key linchpin issues were solved and then it was, you know, unlocks, I mean, xenotransplantation where you were able to make organs for transplantation in animals that enable then, you know, to have a sufficient number of transplantable organs for human beings. I think we're going to probably get there in the next 10, 10 to. Interesting. So regenerative medicine makes a like a real comeback. I think, I mean, I think, well, I think another area, yes. I think on zero transplantation being one, I think the other is going to be, we are going to start to solve problems of regenerating tissue. We already see examples where we, in our own labs,

where we can start to crack, how can you regenerate cartilage, or how can you regenerate other tissues in the body, which would again seem like science fiction, but I think actually harnessing the pathways to really get regeneration to happen, which would help healthy aging is another thing, I think, will likely come.

So there's a lot of things that are still on the way. Can you imagine a moment in time where aging becomes a therapeutic area for pharma companies? You know, we had actually an aging program, a small aging program for some time where we were trying to work on things like sarcopenia, which is muscle wasting and similar kinds of conditions. It turns out to be very, very difficult because, again, multifactorial. And you probably need a medicine with behavior, with diet, with exercise, with all kinds of things to actually. help healthy aging happen. But like I said, I mean, we continue to focus on more the pure regenerative parts. I mean, if you think about the whole world of joints and movement has not really been addressed and cracked. And so this is an area where we have exploratory programs to see,

maybe we could find something. I mean, if you could regenerate cartilage or tendons or enable muscle strength incrementally, you might be able to improve a healthy age in quite a bit. Fabulous. Why don't we actually shift into the innovative medicines set of therapies? Another big area, hot area is in the world of RNAs. So these are really ways to deliver, let's call it, genetic instructions into specific cells. This has been an area that's been worked on for many years. It's always been difficult. But I think companies are now starting to crack the problem of delivering RNAs into specific cells in a highly effective way. Can you give me just a concrete example of how that plays out with like a real disease? So there's a couple of really nice examples now with RNA interference. One that our company is working on is RNA interference to impact a factor that's really a big part of heart disease. It's called LP little A.

LP little A is actually thought to be one of the remaining risk factors for heart disease that have not been addressed. You know cholesterol, everybody's, of course, addressed cholesterol extremely well, triglycerides. LP little A is another factor, but there's never been a medicine against it. And it turns out it's really hard to drug LP, LP, LP, little A. And so the only way to really target, it turns out to be using RNA-based therapies. These RNA-based therapies are able to block the production of the gene, a translation of the gene into the protein and then actually reduce the LP little A in the blood. And so this is one example of how we're trying to take this into an area where otherwise you wouldn't

necessarily have a therapeutic against something that could have a big impact for patients with with prior heart conditions. So RNA interference is essentially a mute button for a gene of interest. That's right, yeah. I love that. That's a great example. And by the way, L.P. Little A sounds like a name of a wrapper. And just remind me really quickly, like, you know, obviously I know what I learned about

RNA from like biology class in the sense of proteins. But can you give us a little bit more distinction about what's unique about RNA-based therapeutic modalities? Absolutely. So when you think about the history of our industry, maybe another way to describe the trend I see that's happening is we used to be about chemicals, the small molecules. So for probably 100 years, I mean, most of the pharmaceutical companies had their basis in the chemicals industry.

And so we made these small molecules that happened to have various effects on the body. And over 100 years, we figured out we could really target what those chemicals do. Around the late 1980s, we realized you could actually make large molecules, large proteins, and make them be therapeutic. So this is antibodies and recombinant proteins. And that led to a whole new renaissance in our industry. And so over the next 20 years and up to today, probably still the largest category is so-called biologic medicines.

These are antibodies and proteins. What I see happening now is a shift to a next set of modalities that move beyond small molecules and proteins. And that is now really touching other elements of what happens in a cell. So one is RNAs, which is really the way DNAs. gets translated into a protein, it goes through an RNA. So that's one new modality. Another modality, both of them really are about editing DNA in different ways.

One is to take the cells out of the body and edit the DNA of the cell or enable the cell to produce something different. The other is to do it inside the body. That's what we call gene therapy. So we make that distinction of cell therapy and gene therapy. So cell therapy is ex-vivo, gene therapy, inside outside. Inside outside. So these are new ways of actually delivering medicines or creating medicines in the human body.

And now you see early stage companies doing even more radical things, trying to turn red blood cells into therapeutics amongst other things. So it's really an expansion, if you think about it, of the game board of how you can address human diseases. I love sort of the sweeping history you have here in terms of starting with chemistry and then moving to large molecules and then now moving more into the cell and engineered world. Historically, every single sort of drug program has been a very bespoke thing, a very sort of, you know, its own ground war, right? You have your target discovery, and then you have your validation,

and then you have your lead, and then you optimize that molecule, and then so on and so on. And at least my sense has always been that because it's so bespoke, that there are some learnings that are generalizable in any given disease area, but every sort of program is a unique thing. When you start to move to the RNA world, to the cell world, to the gene world, is it going to become much more of a modular world where, you know, the first version of a cartee is going to be, by definition, less sophisticated than the second version.

But the second version will be built off the first. And you go from being in a bespoke world to going much more into sort of an iterative world. Unfortunately, in our industry, it's always the answer is it depends. I think in the specific example of CAR-T, I do think that's what's going to happen because you have such a complex manufacturing that you're going to have the first generation, let's say, have a CD-19 card, which is a cart that targets B-cell cancers, and you're going to try to then move into a next generation that hopefully has more rapid manufacturing, maybe higher efficacy, and then even more rapid manufacturing. So you're going to get into that iteration. Now, it's not like medical device iteration. I mean, this is still going to take years to do, but you are going to get to that iteration. I think another way, what I see happening, though, with these new technologies is real platforms insofar is once you have the backbone of the production and even the go-to-market model depending, you can put multiple products onto the platform.

What we've done at our company is build a global network of manufacturing sites that can take cells out of human beings and reprogram the cells and put them back in the body. And we've built the links into hospitals to enable us to do that. So you have that as a capability. You also have the capability to understand how to use what's called a lentie virus to reprogram a cell. So we've got all of that. Now we can apply that in very different ways, in cancer and sickle cell disease and inherited disorders and use that same infrastructure to actually then keep pushing the medicines through.

That's very different than what we've had to do in the past, where every single medicine had a bespoke production process, had to have its own manufacturing facility. now we can actually build that platform and then layer of medicines on. It's no different in gene therapies. When you think about AAV vectors, these are ways to deliver these gene therapies into the body. Once you solve it, the process, let's say, for one of these vectors, you can apply it to multiple different diseases and not have to recreate everything again. That's a shift I see in how our industry operates. You know, I find that fascinating because it actually sounds a lot like what we talk a lot about around this theme around engineering biology.

and when you bring engineering principles and mindsets to biology. You've just mentioned multiple places where there are sort of repeatability and different aspects of engineering have already come in. How is this trend going to continue? Where are they going to be the new places where engineering can play a role? I think the easiest place is going to be in continuing to innovate on the processes by which we really manipulate cells and gene and really get to the next wave of manufacturing. Because I would say we're really on.

the only learning to crawl with respect to most of these technologies and how we produce them, pretty rudimentary. And so I think there's going to be an engineering problem of how do you handle cells and how do you handle the vectors and make this a much, much more efficient process. And there's a lot of, I think, very smart engineering firms now working on that space. So I think that's one place. The area I'm quite interested in is how we can get much smarter at actually engineering the medicines themselves. I mean, we spend a lot of work investing in AI and 3D visualizations to say in the so-called world of chemical biology or if you even think about using quantum chemistry that really understand how to how to define your monoclonal

antibody. How can we do a lot more engineering of medicines up front? Because we really come from a heritage where everything was just trial and error. We just tried many, many, many molecules until we found one that worked and we just took it forward. Right. How can we become much smarter about that. And so in our research labs, we're spending a lot of time thinking about how do we engineer the medicine up front to do what we want it to do. And that's a whole new world, I think. Yeah. Also, I think there's going to presumably have to be a culture that shifts along with this. I read Alan Greenspan's book of The History of Capitalism. And he talked about how actually in Europe, like furniture was bespoke and you'd make this beautiful chair and it's

this handicraft. And they actually hated the idea of factories. The engineering because, you know, it takes the art out of it. It's not artisanal. Yeah, it's not artisanal anymore. But I think once you can have this ability to shift towards that mindset where you have reproducibility and like almost like a factory-like process that can be built, once you can have that shift, as long as everyone is ready to make that shift, then things can really start rolling.

But there has to be a major shift. In terms of in America, people didn't really care about the artisanal part as much. And we got factories. And that was a huge part of the early late 1800s. And I'm curious, you spoke so much about how the virus is changing. And so presumably there's an internal cultural change as well. Yeah. We're trying to make a quantum change, I think, in our culture.

I mean, what we have is as context, I believe, we've moved to become truly just a knowledge organization. I mean, so much of the rudimentary tasks have been either automated or sent to third parties. So we have a whole organization of knowledge work. Well, 50% of them are millennials. And they want to work in a very different environment. and let's say an industrial company 20 years ago. And so we call our new culture inspired, curious, and unbossed. And we want our people to feel inspired by the work, really curious about the outside world

and not lived in a bossed company, but really live in an unbossed, much more empowered company. And when we talk about areas like digital and data science, sell, and gene therapies, it's so critical because these are so complex areas. You need your people to figure out the answers. And we can't be in a world where everybody's waiting for man. to tell everybody what to do because none of us know what to do either because these are a whole new new spaces for us. So that's a big shift. The other element of that journey is to get a lot more

comfortable with rapid failure. I mean, we have to be much more rapid cycle. We can't expect that we're going to sort it all out and it's all going to work perfectly because the first thing we've learned already in cell and gene therapy is nothing works the way you expect it to work. And so you built a platform for rapid generation. That's the idea. Yeah. What I love about that is it reminds me of computer. software companies and the shift from waterfall to like more DevOps, agile, even microservices architecture. We're a little late to the party, but yes, that's the idea.

It's the same kind of principle. That's fascinating. So we haven't talked about the big elephants in a good way in the room of AI and ML, you know, artificial intelligence and machine learning. Let's talk about AI and ML and data. I mean, it's not a question of if, when, it's how. The question I have, because quite frankly, it's a very hype topic too. and people sort of promise all kinds of things

when they talk about applying AI and ML to medicine. I'm very curious from your take as a head of Novartis, where do you see the strongest applications of AI and ML? Well, I have to first say, I completely agree about the hype cycle here. I mean, as we've gotten quite scaled and working on digital health and data science, we've learned that there's a lot of talk

and very little in terms of actual delivery of impact. But we've learned a lot. I think the first thing we've learned is the importance of having outstanding data to actually base your ML on. And in our own hands, in our own shop, we've been working on a few big, big

projects. And we've had to spend most of the time just cleaning the data sets before you can even run the algorithm. That's taken us years just to clean the data sets. And I think people underestimate how little clean data there is out there and how hard it is to clean and link it was never intended to have this type of analysis

done, right? It was intended to, for, a given project and that's it. Yeah, that's been so much of it. And then the other thing is, is there patterns that can be really learned from the day? I mean, do you have a good training data set to actually train the algorithms? So there's a few places I think we've seen a lot of traction. One, I think the vision or image problem has been very well, well solved. So right now we're in the process of digitizing all of our pathology images and having AI just be able to scan all of the pathology images at Novartis, and we have millions of, of course, records of biopsies and tissue. So that's a huge project we have called Path AI, really work on that as a single

example. I mean, that's like a gold mine. It should be. I mean, it should be. And if you then apply that as well to the vast stores of imaging data we have from our clinical trials, we have two million patients in clinical trials, at least in the last 10 years. And we have MRI, CT scans, retinal scans, heart scans, and all of that as well, I think ML can have at least a significant potential to really find hopefully new insights. I think the vision image problem has been one we've been able to really take on. Another area is in our operation. So we build an operational command center. Take us, as I said, two and a half years to build it. We call it sense. And what it enables us to do a team sitting centrally in our headquarters to look at all of our clinical trials in the

world. And AI is predicting which trials are going to enroll on time or not enroll on time, predict which ones are going to have quality issues or not quality issues. And the reason we could do that is we had 10 years of history to train the algorithms. And we run about 400 to 500 to 500 clinical trials a year. So we have a lot of data that we could train the algorithms. Does that mean you've had to dig all the way back into automating sort of real-time information on clinical trials? So the data entry on a clinical trials? So the data entry on a clinical trial as a patient is enrolling? Is that all been automated as well? Because that used to be done on pads. It's a great question. I mean, really what we focus on is the operational data.

So one level up from the patients is the trial enrolling on time. Are the sites open? All of all of that, all of those elements. On the operational side, it was really easier to do this than trying to get all the way down to patient level data. The other area, interestingly, in the financial area as well, we find that AI does a great job predicting our free cash flow, predicting a lot of a lot of our sales for key products. And it does better than our internal people because it doesn't have the biases. And the data is very clean. And we've got very long-term data. So that's been all positive. But there have been other areas where I think it's just simply not met up. I mean, I think the holy grail of kind of having unstructured machine learning go into big clinical data lakes

and then suddenly find new insights, we've not been able to crack mostly because the data to link it up. I mean, we are spending a lot of our energy just trying to get all of our data harmonized so that some algorithm could maybe find anything of use. There's an area that's desperately in need, I think, of innovation is how we think about clinical trials, recognizing we have to operate within the system that we live in. But if you could design testing safety and efficacy in humans on a blank sheet of paper, what would look different from a clinical trial perspective versus where we are? today and the way we do it now.

I mean, the ideal world, if we could ever, if we could get there, would be we would have integrated health records where we could easily insert the fields that we needed for clinical trials. And then we could use something like a blockchain or some other distributed architecture that enabled patients to consent for us then to access the data and then, you know, run the trials through that. And that would eliminate so much of the effort of creating a, second database versus the EHR, monitoring that database, QAing that database, locking that

database. You could get the data on an ongoing basis. I mean, we would radically, radically, you know, simplify this. I believe that's a huge, huge opportunity. I think we have a long way to go, as you know, because EHRs are not where they need to be. We're probably not where we need to be to get there. But I see opportunities in baby steps to actually get towards that. And I think we're experimenting with that. I think other companies are as well. The other thing people talk about, but I mean, I'll take a skeptical voice around it,

is the ability to use real-world evidence to try to get at these things. But as somebody who's worked in clinical trials from most of their time in the industry, I do believe that, you know, the power of randomization, the power of blindedness is what enables us to control for all of the things we don't know about the complexity of human life

and human biology. And to think that we're going to take that away and then be able to really determine the efficacy of a medicine puts a lot on the statistics that I don't think we have. And so I'm more of a real-world evidence, I don't know if it's a skeptic, but

realist who sort of says, after we have randomized placebo control data that really tells us that something has the effect we think it is, then to explore more effects or explore more uses through real-world evidence makes a lot of sense. But I don't see this as a panacea

that suddenly will make the world much easier. Yeah, I mean, that's my expectation as well as you'll see it first come out like as a phase four. You know, something where, you know, you're using real world evidence, which is right now used for reimbursement anyways and so on. And but then maybe see how far I can go back, but it's not going to replace it.

You guys don't think a secular, I mean, not to sound naive, but you don't think a secular shift like censorification of everything and everyone really truly has continuous wearables. Like, everyone's wearing a CGM by default. I hear you on the statistical side, and there's a lot of other spurious variables and things introduced into that equation.

But it is a huge, it's a very deep, nuanced, patient-level set of data that seems like we can't ignore the power of that. Like, where do you fall on that? When I think about, first of all, I would say just in general, in sensors, is another place where there's been a lot of hype above what of expectations.

I mean, we've been really trying to explore the use of sensors in clinical trials now for, in my own experience, at least six years. And it's been tough to get sensors that really meet clinical trial grade outcomes. I mean, to really show that they can be validated versus our current clinical endpoints. Now, if it's consumer products, fine. I mean, perfect, perfect people can build. But here, we need to really be able to replace what are pretty rigorous tests. And we haven't seen that yet.

Now, we're exploring, I think, use of many different sensors. The real power of it is a continuous variable to actually see how a patient's doing in between the study. Between the study visits. And so I think that will help a lot. But I still think in the end, you're going to need to randomize and blind. I mean, I think if you don't randomize, I think it's really hard to figure out what is going on in a complex system. I agree with short term. I think longer term, my good feeling is that this is a solvable.

problem statistically, because there is even issues with clinical trial design that one has to overcome today because randomization isn't just picking people literally randomly, you know, necessarily. It's a sample, not a population. And there's been a lot of work on causality theory and statistics that have come around. So there are advances, but I think it's not there now. Yes, I agree. Small N, not capital N.

More to say there? That was really interesting. What's the role of bringing innovation in from the outside through partnerships and M&A and ML? Yeah, I think one of the things we're working through is how do we get the talent, you know, as we really start to organize the data and we've brought in some great talent to really help us work on data architecture and come up with a whole data landscape for the company. So that we're always now thinking about how do we treat data as an asset. That's one of the things we keep harping on is data as an asset. Whatever data we collect from the external world has to be organized in a clear data architecture. But then to take the next step to get the data scientists to really find the insights. We're not the traditional place where data science is coming out of Stanford is looking for where they want to come to. So we're working through partnerships with universities, potential partnerships with startups. Actually here in the Bay Area, we have a center called the biome where we're working with different startups.

And so these are the things we're trying to do to engage and hopefully create an ecosystem that helps us do this and not just do it ourselves. I don't think we'll be able to track the scale that you would need. Yeah, there's a Riesis peanut butter cup issue because startups sometimes have some innovation on the data science, but not the data. And so bringing the two together, I think, seems like a very natural combination.

Where does it mean? Peanut butter and chocolate. Like, she's got the peanut butter. Oh, my God. I'm like that. You don't remember those commercials? I don't remember them. I watched a lot of TV knows growing up, but I don't remember that.

I find it fascinating because a lot of our bio entrepreneurs, the number one thing that they tell me that drawing data scientists, the bio companies, one of the hardest challenges they have to face. And so you're saying with the biome and other things that you're doing that you're essentially saying you have to kind of create the pipeline, not just source it. That's right. That's right.

And to what you're your earlier point, I mean, the opportunity is to say, look, come and work with us and we'll let you work with our data and you can learn and we'll learn. And maybe then there's a partnership that's created or maybe you want to come work for us, which would also be great. But that's how we're approaching it. Well, and there's actually an interesting shift that can happen in academia

with my group at Stanford, many people actually, during their PhD, have gone to work in pharma. And it's hard to, it's impossible to pull the data out of pharma, but it's actually easier to put the grad student into pharma. And so the grad student comes with the code,

runs it internal through the firewall of pharma, and we see how it does. And you can still publish papers where maybe you have to obscure what the target is or something like that, but you can at least see how things are going. And there's nothing like sort of trying it in the real world.

Yeah, yeah, makes total sense. So on this question of bringing in talent, so you guys operate globally. Obviously, you're in 150 countries somewhat. You're headquartered in Switzerland, Nibers in the Boston, Cambridge area. You have a presence out here in Silicon Valley. So how do you guys think about innovation hubs? Very simplistically, is all of the machine learning, artificial intelligence, talent going to be based out here? How do you sort of distribute teams across the world? So it's interesting. You know, when you look at research, we have three, main hubs, our three main hubs are in Cambridge, in Basel, Switzerland, and in Shanghai in China. Those are our three main research hubs. In terms of development centers for product development, you would add on to that list, Hyderabad, India, as kind of the main, and East

Hanover, New Jersey. But when we comes to data science and digital, what we've actually decided to do is take a much more distributed approach. So we're building up these biome centers in San Francisco, in London, other locations in the Middle East, perhaps in China, just trying to say we're not going to constrain ourselves with our current locations. We're going to just try to source talent wherever it is, particularly because talent in these areas doesn't necessarily have to be housed next to the other functions. We're really asking these people to explore our data and find big, big new insights. So that's the approach we're taking right now.

It was really saying, you know, let's go where the talent is as opposed to force everyone to come to us. So we'll see. That's the experiment we're undertaking. How do you see the future of that sort of working out? Do you see that Boston, Silicon Valley, Basel, like these places will specialize? Will they distribute? Yeah. We have lots of debates. If we were to build a scaled hub in digital or in data science health, where would we go?

I think one of the challenges in the Bay Area is, again, just the competition for talent is so intense, especially in the tech sector. So we're in the business of funding early stage companies, supporting. entrepreneurs. If I'm an entrepreneur, I obviously see a ton of benefit in partnering with Novartis, access to data that doesn't exist elsewhere, obviously validation in my approach and my technology, et cetera. But if I'm an entrepreneur, I'm also scared to approach a large company like a Novartis because I'd worry about, you know, basically you're an elephant and I'm a mouse and if I want to dance, I have to hope you're a very graceful elephant. Otherwise, you're going to crush me. What advice would you give to entrepreneurs about approaching biopharm, a large bioform

in the spirit of collaboration? Yeah, I think in data and digital, what we've tried to do is make us feel a lot smaller because I think we recognize that we are a huge beast. And so with things like the biome, we work with many other entities to try to say, how can we make ourselves feel smaller, work in smaller units, we created our own digital data organization so that entrepreneurs would have an input into Novartis where it's people like them. I mean, the people in that team are all come from the tech sector. They are working in a much smaller, agile way. They do sprints and scrums, and they work in all the ways that the people are being used to working.

And so I would say really engaging through someplace in a large company that I think has a natural affiliation for the entrepreneur makes a lot of sense. I think it is harder on the kind of traditional biomedical side, right? I mean, we have, I mean, if you just think of, we have 17,000 R&D people and spend $9 billion plus a year in R&D. So if you're a small entrepreneur who wants to start working with us, it's easy to get lost in the fray. We're trying to work on that. I think most of the companies in our industry try to have external offices that try to engage. I mean, we have external scholars program where we really try to enable.

able scientists to use our facilities, interact with our scientists. So we're trying to experiment, but I can't say that we've completely figured that out on the biomedical side. I'm much more optimistic on the data and digital science side, mostly because we just brought people in from that world and they just think differently. There was something I wanted to ask you earlier, which was about measurement. Because when you talked about the portfolio approach, I wanted to know how you think about actually measuring the way you make those investments in a portfolio. And the reason I ask is because, you know, there's all these mindsets like pastures quadrant. Like, here's a place where we're going to put more emphasis on basic research and we're going to put

more emphasis on something more practical. Or there's another approach in Xerox Park. They used modified real options analyses as a way to figure out how to do like short term, long term, midterm type investments. Do you have a way of sort of closing the feedback loop for how you measure the success of how you're allocating and deploying investments in R&D? Yeah, I mean, we have financial measures. So we look at return. on capital employed, NPV, NPV, peak sales, all the traditional financial measures. We look at really the scientific innovativeness,

for lack of a better word. Is this really something that's changing the game from a scientific standpoint? That's a little bit more of a subjective measure, but we try to ask teams, is this really moving the needle from a standard of care, science, and we actually score that based on six different parameters? Oh, interesting.

Are you allowed to share those parameters? I don't know them off the top of my head, But, you know, we really try to score the, score the medicines to say, is this really transformative? So you have a financial score, you have a transfer, transformational score. And then another kind of subjective element is, does this strategically fit? So is it in one of our core therapeutic areas? So if somebody comes with a great breakthrough, which happens not quite often, in an area that we're not in, that's the toughest one because it's a big breakthrough, but we're not in this space.

And what do we do now, right? And do we really want to build this up or do we want to just send it to an outlicensed to a fund or do something else? Those are tough discussions. But we try to be disciplined because it's, again, the patience and being really sure you build depth in your key areas. Because if you take another program on, that means there's another program you have to stop. I mean, it's a zero-sum game for us. One thing that's funny, just listening to you talk about what Sonal brought up this question of not invented here syndrome. and when you contrast that with managing, you know,

having an organization that is naturally curious and unbossed, as you said. Inspired. But managing that non-invented here syndrome versus maintaining sort of the skepticism that things might be in a hype cycle and not sort of chasing hype, it's a very fine balance, right? It's kind of like the not invented here, the other side of that coin is not invented yet. And you've got to figure out like where you are in that.

And I think that is one of the most difficult things that I would imagine that an innovative company at this scale at which Novartis operates has to always find that balance between. Absolutely. I mean, there is a balancing act between the different forces. And I find a lot of it comes down to just encouraging people just to have open, frank debate and be comfortable with task conflict without personal conflict. That's what I keep telling our team. We have to be incredibly curious about what one another thinks. I think that's just all about trying to get the best ideas, and we're just trying to debate, but it's never personal.

And it's never, because I think when, particularly in the world of science, it often becomes personal, right? It becomes this is about me and my science versus you not believing in my science, as opposed to saying, we need to just find a great medicine or we need to just solve this problem. That's a journey, I think we're taking the organization on. But I think that's going to be what's really critical is having that radical transparency in the open debate. I find it fascinating because it alludes. the concepts around skin in the game

because you want people to have skin in the game but at the same time they need to have just enough out that they can see things a little clearly where you're not like only attacking their sacred cows.

You're skin in the game but not vital organs. Exactly. That's a great way of winning it. I love that. How long have you been in the CEO's chair now? One year. What's the, you know, having come up through the R&D side

of the organization, what's been the most surprising thing to you now as the CEO given that R&D is such a big part of what the company does? I'm just, I am just amazed.

is by how vast our company is. I mean, you know, I think even though I've been at the company since 2005, now actually overseeing a company that's 120,000 people and 150 countries, and you go anywhere, we are just a vast, vast company. So that's one thing that's really, I think, surprised me just to have to now, when you think about making a transformation happen and you try to make that happen in such a large enterprise, that certainly really, I mean, that really hits you. I think the other thing about this job is crisis management, which, you know, you're just not exposed to.

I mean, this job is a lot about managing crises, and that's been a big learning curve for me. Because in the world of R&D, we had clinical trials the last two or three years. I mean, everything's sort of predictable. We sort of know what the decisions we need to make. A lot of documentation that you can lean on. Now you're in the world of the ambiguous, the uncertain, and then things hit you from, you know, completely from the blind side. and then you got to keep moving ahead. If you were to write a letter to grad students

or just people kind of entering the space, like what kind of skills would you encourage them to have? Like if you could have added things 20 years ago, what would you tell them to do? I'd say focus a lot on how you lead people. I think there's so much of a focus on technical expertise and thinking that that's going to get you there. It matters, of course.

Competence matters tremendously. But what really makes the difference is how you lead people, how you lead yourself. And I think investing more in that, would pay off a lot. I think the other thing I'd say is don't underestimate the importance of getting multidisciplinary exposure. I mean, I think most people get worried when they have to make those jumps. I've had a career at Novartis where I've worked in commercial areas and marketing areas, so most of my time in R&D worked across four different areas of the business. And so with that

diversity of experiences, it enables you, I think, to take the right decisions. There was one other point. I wanted to raise. I think what's often lost some people, because you mentioned the miracles, right? And how incredible it is that we find any human medicines at all. Because if you think about it, every human being is probably 40 trillion cells that are

working together. It's amazing. Anything even works. It's amazing. We understand a fraction of the proteins, what they do, 1,200 drugable proteins, and there's only a fraction of those that we can actually drug. We don't know what most of RNA does, non-coding RNA. We don't

know most of what the genome is even talking about. And if you look at it, since the creation of the FDA, there's only been about 1,500 new molecular entities ever found. Wow. And most of those are actually overlapping in similar therapeutic areas. So actually, if you were to count for, I haven't done the analysis. But if you count for double counts, my guess is it's in the hundreds of medicines that we've actually found. And by the way, what's a predominant therapeutic area? Probably, I would guess, hypertension, cardiovascular disease, but I've not looked carefully. But it's worth reflecting on how hard it is to do what we do. And when we find, I tell our people, you have to think every medicine we find is a miracle that fits in the palm of your hand.

We've unlocked, in a sense, a billion years of evolution of the eukaryotic cell in human biology. And somehow we found something that was able to move the needle in this incredibly complex system. I think that's easy to forget when we just kind of overly simplify what we do. That's a great note to end on. Voss, thank you for joining the A6 and Z podcast. Thank you. Thanks so much.