René Hoet
Taking risks and driving innovation in biopharma

Nicola: René, thanks for being here. During your 25 years of career, you have worked with organizations of all sizes, small and large, and you have seen technologies come and go. But which one in your experience has been a transformative technology that really changed the game in drug discovery?

René Hoet:

It’s a very good question. It’s also a difficult question. Because, if you go back, I think now 35 years that I’m working in drug discovery. It’s always easier to see, you know, from development, which were 20 years ago, than from the ones which I’ve worked on a few years ago, because it takes time to develop something.

But if I think about the technology, which actually enabled others to develop drugs, I think my work at Dyax actually helped a lot of companies to develop drugs. Nowadays, it’s actually technology, which is well known as semi-synthetic phage library. And we actually licensed that to over 20 different companies, pharma and biotechs, and used it ourselves.

And from that library, there are five antibodies that are now approved by FDA and in the EU. This is published in 2005, and there’re still actually companies working with that library right now. At the moment there are hundreds of libraries and technologies, and it has grown but it was one of the first antibody libraries that was successful.

So that was technology-wise. If I think about a company where I was most impressed during the years, I would choose GenMab, where not so much technology wise, but therapeutically wise, they have been continuously innovating in immunology, bispecifics, they developed their own platform, DuoBody Platform. So they have developed many drugs that are all in the market right now. And although I worked there only for two years, I learned a lot there. And also that company, I think has contributed a lot to drug discovery in the last 25 years.

Nicola: So, you mentioned two companies, one I guess was a bit smaller and one a bit larger. Do you see a difference in which these different-size companies relate to new technologies, whether they might be risk averse to try something or did you notice anything in how they approach new technologies?

René Hoet:

Yeah, I mean, GenMab also started small. It’s now a big company, but at that time it was not so big. So I think it was always driven by innovation and the drive to succeed in the product. Dyax was indeed a bit smaller and it was earlier in new technology development. I would say it’s probably easier for a small company to have that drive and passion and you can influence technologies easier than with a large company.

I also worked at Bayer. I learned a lot there, what it takes to develop a drug. But it’s more difficult to influence a company of 100,000 people, you know, that’s not so easily done. And that’s why I decided in my career to step back from these big companies, going to smaller biotechs again. Because there you actually have less people, you can do less, but you can directly influence what you want to do.

Nicola: And do you think that in that respect, a smaller company can be more innovative because maybe it’s easier to embrace new technologies? Or do you think that also maybe the capacity of large company is more enabling in terms of innovation? What aspect do you think is more favorable for innovation?

René Hoet:

It’s probably true that most of the smaller companies and biotechs have more innovation than large pharma companies. The large pharma companies very often either they buy or they collaborate with the small biotechs to get innovation in their company. So I wouldn’t say there is no innovation in bigger companies, but it’s more via collaboration very often.

The last few years, I worked as a CSO for immuno-oncology drugs, very innovative drugs. That is a real innovation, but I cannot tell you yet if they will be successful because they’re still in clinical trial. So it’s too early to tell if these concepts will be successful, but it’s certainly highly innovative.

Nicola: I can definitely imagine that innovation works differently in these different companies. You mentioned you worked with phage display at Dyax and that it was a major improvement back in the days. When working on that, what do you think was the key aspect of that innovation that really drove the technology or the field forward?

René Hoet:

It’s a good question. At that time, there were a few companies working on phage display libraries. There was Cambridge Antibody Technology and MorphoSys as two big companies. They were having different approaches. For instance, Cambridge Antibody Technology didn’t give access to other companies to develop the technology and that was a hurdle. There were patents actually blocking many companies to develop it. So what they did in Dyax, we developed a new technology which actually was to replace PCR (which sounds a bit strange), but that actually was a new technology to make new antibody libraries. And when that was successful that was a breakthrough because we could license this technology to many pharma and biotech companies. So, it was scientifically but also commercially a success because of that.

Nicola: So what you’re basically saying, if I understand correctly, is that opening up the technology actually helps putting the technology forward and extracting most value from the technology from multiple parties that might benefit of such a new technology.

And I think we discussed also before in previous conversations about the fact that you also have been at the edge of computational innovations in some of the companies that you have been working with. So where do you think that – specifically at the intersection between indeed biology and computation – you have seen the most innovation? What is driving in your experience the largest innovation at that specific intersection?

René Hoet:

Both at Dyax and Bayer, we actually were using quite a bit of automation in our drug discovery in collecting really at that point already sequencing, CDRs, frameworks, antibody sequences, hundreds of thousands of sequences. At that time, I’m talking about Dyax here was around 2004-2005. So this was pretty early on. There was no NGS at that point yet. In Bayer, there was NGS, that increased the number of sequences. But we started early on with collecting sequences and collecting binding data and cell binding data, affinities, and put them all in the database.

At Dyax, we had bioinformaticians that actually worked on a program we called Webphage. In Bayer, we actually worked with Genedata as one of the first companies to actually start making a database platform, which they later expanded and licensed to many pharma and biotech companies. So that’s sort of my involvement. I am for sure not a bioinformatician, but I understand the major questions which are important.

That’s a bit beyond what’s now going on with AI, I would say that’s a different story. But it started with good collection of good data, I would say.

Nicola: So, basically what you experienced is the increased amount of data and the necessity to be able to organize that data and put it in the right context. And you mentioned AI and, of course, this is a very hot topic. What is your take on that? Where do you think AI is really improving currently the field and where you see it going in the near future?

René Hoet:

I’m very intrigued by AI and all this fast development. I mean it goes so fast. If you don’t look for a few months there are already new models and new approaches available. So, it actually is very fascinating to see. On the other hand, there are also some people that are already quite disappointed by AI because they expect that AI could do everything.

It’s probably true with any new technology that there is a huge enthusiasm and then there is some disappointment. And then you get some steady flow of that it’s going to be applied and really useful for many people. It is already making a big impact, and it will be much more even in the future.

I think it’s important to test the different models that are available, collect the data, test different models. At the moment, for instance, in the antibody drug discovery, it’s already pretty good for optimization of molecules to use AI. I think most people are convinced that’s actually the case. But for discovery in silico design of new antibodies, it’s still early days, I would say.

In that context, I found it interesting that a colleague of mine, Andrew Bradbury, set up his contest for many companies to evaluate different antibodies that had the COVID-19. And he presented the result quite recently, and it was actually interesting. On one hand, it shows that AI works to some extent. On the other hand, it also shows, compared to his own experimental data, that most of the AI-designed antibodies were not as good as experimental antibodies. And only in a few cases, they were a bit better, a few antibodies were a bit better. So, there’s still a lot to gain there.

And what I find also interesting from that study is that there’s so many different models and everybody used different models and it was not clear which model was best and very many of the companies used a combination of different models so it’s not so black and white yet.

Nicola: So do you think also based on that study – and I personally believe that more of these comparison studies are necessary to understand where we are currently in the cycle, let’s say on the AI cycle of technology development – do you think that in this cycle, we are still climbing the hype – let’s say mountain – and we’re not yet at the top? Or you think we are reaching the disillusionment moment if there is a disillusionment moment for AI in biopharma?

René Hoet:

Yeah, it’s a good question. I don’t know yet, I think it’s still in a bit hype phase, I would say. But I think big companies particularly find it difficult to implement this technology in their workflow. There’s already maybe a bit of disappointment from higher management that it doesn’t go as fast as they wanted to see it go. So, it’s probably there, sort of at the top and we’ll see where it goes. Hopefully, it could also be a technology which is so fastly implemented that there is no big disappointment, but I guess there will be a bit.

It’s probably also due to that some of these AI companies promise to change the world. They say they can design antibodies in silico, that they only have to test three antibodies and they got a picomolar affinity antibody, which I think most people in the field doubt this is correct. And these types of hype create for people a little bit outside the field the idea that this is all easily done, which is not. And I think a company of ENPICOM is very important to really support companies in the development of different AI models and the implementation of it, and to compare these different models.

Nicola: Yeah, and indeed, I think there are many claims out there – quite bold – and I think those have the potential to create some disillusionment or some discontent if those don’t actually match eventually. So, we are still climbing a bit that high mountain, but there is that potential of being disillusioned by it. And regarding this, I think, again, you have a very long history, and maybe this is bit provocative question, but have you seen any technology in your career that you had great hopes for, but then eventually turned out not to be that great when you started implementing it or you to dive into it?

René Hoet:

Yes, quite a number, I would say, because you try new things and you have the expectation that this is easily done, and some things don’t work out as you hope for. Usually, these things are not published if they don’t work at all, I have to tell. So, one of these things which I think I still have good hopes for (there are papers about it, but people still question how efficient is) is phenotypic screening. People typically take a target and then do target discovery based on that target and optimize the antibodies. For small molecules, it’s actually quite common to use phenotypic screening, hey start with a particular assay and then they screen the whole big small molecule library for certain hits and they find some initial hits which they optimize. So, they start with the function first.

With biologics, it’s the other way around, which has the advantage that you know where you’re going, but the disadvantage that you get maximum what you expect. So you don’t get new things. Real innovation comes from things that you don’t expect. And there I feel that phenotypic screening could actually make a difference.

For example, if you start off with cancer cells and you select certain cancer cells and you deplete on normal cells and then you screen for some function of those to inhibit proliferation, then you can find quite novel things. For example, novel glycosylations on cancer cells, it’s known that they are there. So you could select for a novel glycosylation site, which is only on cancer cells and not on normal cells. And people have been successful in that approach. So there are even antibodies found early on. And one of them is, for example, alemtuzumab, that’s an antibody called Campath, which was validated in the clinic already in 1984—so a long time ago—without that the target was known. And only in 1991 the target was identified, so there are historical cases where people actually found that. I would say in the last 20 years, this is limited. There are a few companies that actually are doing this phenotypic screening.

For example, BioInvent, they have this platform function first. They screen antibodies first for function and then characterize the targets and they have some antibodies in clinical trial. And I think the advantage of this function first is also that you can easily incorporate the basis of the patient, you know, use patient data. I think that’s for me the important step. If you start with a target, then you only at a later stage screen typically for the function. You find first antibodies that bind and maybe the affinity and then you go to function and then in the end you look at what impact that has and then you’re typically limited to a few only. Well, if you first select, for example, an agonistic function in high throughput and then you go into the next stage, then you come from a completely different angle. So I still have good hopes for that.

New technologies might help here. There’s new microfluidic technologies where you can now screen millions of cells. If you can screen millions of cells for a functional activity, then you might be able to come there faster because that was one of the limitations. It was very difficult to screen these millions of cells for a functional activity. In particular, if you think about B cells that are not immortalized. They don’t live for a long time so you have to do the screening in a very short time window to be able to be successful. We are still working on this in FairJourney Biologics with mammalian display. There you make a copy of the B cells in mammalian cells, you incorporate it by stable cloning, and then screening is set up for functional screening. That’s still in a development stage. You can find easily binders to a molecule, but the idea is in the future this will be possible also for functional activity. And I think that might help also to find some novel things.

Nicola: Do you think that the main reason why, for instance, phenotypic screening is being applied more successfully to small molecule rather than biologics and antibodies, is mostly because there was a limiting factor in terms of the number of antibodies that could be screened? Or there were also other factors that you imagine contributed to it not that being more widespread technology for finding new drugs?

René Hoet:

If you start with a cell which has 10,000 different targets and they’re all different glycosylations and you apply an antibody library of 10 variants, you need to be able to screen that in a sufficient way. And the display technologies typically use not full antibody molecules, only fragments, and you cannot test them that well for function. For instance, FAP or single chain or single domain, typically, are not enough to create an agonistic activity. You need the whole molecule, and there is not that many display systems (except mammalian display I would say and yeast display to some extent), where you can make full molecules, make them available, and make very large libraries that you can then screen for activity. B cells is one of the possibilities. Single B cell screening, which has been developed the last 10 years or so extensively and there you could also try to do this or go from the B cells directly to the mammalian cells and then screen for activity. I still have good hopes that these types of technologies will help create more functional activity.

Another area I think where this phenotypic screening is needed and will help quite a bit is in these bispecific and multispecific antibodies because there you put activities together—two, sometimes three activities—with the idea that you want to have functional activities. And if you have 100 antibodies to one and 100 to another one, then you already have 10,000 combinations. And how do you test all these things separately? Typically, it’s too much to test in producing all these proteins. It would be nice if on a cellular level, you could make these molecules in the library and select them in advance, before doing all that work.

Nicola: Do you think that maybe the tendency for large pharma to go after the same targets and therefore have a more target screening approach, rather than that phenotypic broader screen that might actually unravel new mechanism, has something to do with the tendency of choosing a more targeted approach for antibodies rather than a phenotypical more broader screening? There is a political, let’s say, take on that as well.

René Hoet

I would say particular in large pharma companies that there is a tendency to not take too many risks, and try to reduce risk, and be successful with as little risk as possible. For example, I think you mentioned this already, there’re so many antibodies already developed to CD20, 20 different ones, I think are already in the clinic. HER-2, I looked it up, there are 40 different antibodies in the clinic. PD-1: over 50 different antibodies developed by different companies in the clinic – not just for discovery – all with the idea of one is successful, so we make another one and we try a different combination. Which is in my view, a bit of a waste of energy and time, which would be better placed in trying to find something new.

There are examples also that something new completely changes the world. This is not for biologics, but I’m just thinking on the new drugs for obesity and diabetes: Ozempic that actually changed completely the drug discovery for this type of molecules. For this year, I saw that 4 of the 10 best-selling drugs are now these drugs. 60 billion sales per year. And that’s in a matter of two-three years. These complete innovations can completely change the landscape, I would say.

Nicola: And do you see the possibility to combine the phenotypical screening with the targeted screening? Because I totally see how phenotypical screening can unravel new targets perhaps and open up completely different opportunities for patients and also markets. So, do you see the possibility to combine the two?

René Hoet:

Yes, I think it is possible to combine them. And I think it’s important for phenotypic screening to have a very robust screening assay. Ideally have materials available from the patients there, and then unravel by the target pretty early in the discovery process. There are systems available, and we use that at Bayer also to identify the targets, like technology developed by Retrogenix, which is now taken over by Charles River, with which you can screen your antibodies against all human targets, 5,000-6,000 targets. So, once you find an interesting antibody-worked activity, you can quickly go to the target. And then, because this might not be the best antibody, you have a starting point and you can use a traditional way to further optimize and make more antibodies to such a target. So, I definitely think there is a possibility to combine things. And perhaps AI in the future could also combine a bit earlier, maybe even in finding new mechanisms or combinations of targets. That’s an area, I think, which is also very interesting.

Nicola: Yeah, absolutely. That’s indeed one of the questions that I wanted to ask you. Whether the fact that AI might help targeted screening will make it a more sought choice for pharma, because maybe they can actually identify via AI novel target with that technology. So that’s one question. And the other one that I wanted to ask you is: do you think that AI might replace phenotypical screening, and you would be able to do the same type of screening with AI at some point in the future?

René Hoet

I mean, you would use phenotypic screening to create innovation in it, which is probably difficult to predict, at least at the moment, maybe in the future it’s different. So, I don’t see that immediately replaced by AI, but AI can certainly enable it and further develop it.

And the other question you had, sorry, I missed it and I started with the second one.

Nicola: The second one was maybe more provocative, about replacing completely, phenotypical screening. Maybe if we look at the target discovery approach and the fact that maybe AI can help finding new targets and hence, replacing a bit the idea of phenotypical screening as in finding new mechanisms, because that could actually be achieved via AI and then reverting back on that target discovery approach. Would you see that as a concrete possibility or you think you’re doubtful?

René Hoet:

No, I do see it as a concrete possibility. A company where this is already done is a successful company Alchemab, I don’t know if you know that one. They are using patient data from resilient patients who actually recovered from disease, in the area of neurology diseases. They use antibodies from these patients and the diversity of these antibodies, compared with the patients who don’t have this inhibition for disease. And with AI, they identified antibodies that are unique for this patient subset, and then identified also the targets. And a few of those antibodies are now in clinical trial. So I think that’s actually a very nice combination of using AI and the type of phenotypic screening. Different than I mentioned before on cells, but here it’s using directly patient data.

Nicola: That’s very interesting. Where do you see the low-hanging fruit of AI in drug discovery? I’ve been to conferences and places, and AI has been adopted across the spectrum: from both improving molecules, but also in the lab, also in the clinical aspects, extracting data from patients’ reports and so on. Where do you see the most impactful innovation at the moment?

René Hoet:

I wouldn’t be able to answer the things on the clinical trial piece because I’m not really an expert in that one. I see that it can have a huge impact because it’s such a long period and such an important part of the drug discovery – to design clinical trials well for the right patients and look at biomarkers. That’s later on in the discovery, but can have a major impact.

For the early part, if AI can actually design new drugs based on a target structure or structure prediction to a certain epitope, there it could make a huge impact. And perhaps in two to five years, drug discovery for some programs could be obsolete because we know the structure of a target or domain and we can design antibodies that bind to this domain as a starting point and optimize them. That would be a fantastic step forward, from in-silico design based on structures to be done. I think that’s a dream that many people have. I think it can come true, but it will take a few years.

Nicola: Coming back to the mammalian display and these technologies that you also help other companies implement and adopt. Of course, in those companies, there was probably innovation that they were trying to achieve. What are the obstacles that you faced in trying to implement those innovations? Are there specific obstacles that you would have to face in implementing innovation in smaller or larger companies?

René Hoet:

Mammalian display is a quite sophisticated technology, because this is stable integration in the genome and you have to select a huge number of cells to make large libraries. So I wouldn’t say you can implement this easily in any company. There need to be some steps there: molecular biology, cell biology needs to be quite broadly established. For example, at Iontas, which is now part of FairJourney Biologics, they developed the technology and licensed this to BMS. They set up the transfer process and they were able to implement this pretty well, but that’s a big pharma company. I wouldn’t say this is easily done by every biotech company.

Nicola: Indeed, you require a specific set of expertise to do so. Do you think that those are mostly technological issues that need to be overcome or adopted? Or are there also other aspects in implementing innovation, perhaps with management or other aspects like that, that you experienced to be critical to implement innovation?

René Hoet:

Innovation needs the drive to succeed. The passion to succeed is very important. And that passion should come from all levels: investment in equipment, for example. For mammalian display, you need a very sophisticated FAC Sorting machines. So that’s an investment needed. Of course, you can try to set up your own technology that will take even more time. Or you can license, so you need that financial piece to be in place.

And then you need good scientists to implement and that they have time enough to implement and validate. And there should be no unrealistic expectations. That’s also important. I think my experience leads with large pharma companies, they can be very enthusiastic initially, but then they want to see immediately all sorts of results. So, there’s only a small time window you have to succeed.

Nicola: There’s definitely that limitation in time, and indeed you mentioned, for instance, the case of mammalian display. You need to bring together different expertises, cell and molecular biology. Now, of course, with AI there is yet another aspect: the computational biologist, and the computer people joining the crowd. Do you think there is also a cultural aspect to be considered, how to make these different people talk together and be brought together as a key component of innovation?

René Hoet

Absolutely. I think silos are actually the worst scenarios. The bigger the company, the more difficult it is actually to prevent silos. In my career, I’ve always had to try to build bridges between different expertises and structures. I think it’s really important, because that’s the only way to be successful in developing a drug.

An example of company that I have not worked for but that is very successful is the company Argenx, located in Ghent, Belgium. We worked together at Bayer with them, and they did an early discovery program for us. They were not successful, but we saw with all the scientists, the whole company was so motivated and so driven to succeed that I’m actually not surprised that the company is now a very successful biotech pharma company. So, it needs a drive, in my view, from the upper management to the scientists, to succeed.

Nicola: I totally see that the whole organization needs to be aligned towards that objective. And when it comes to bringing the computation into the drug discovery, do you see also the necessity to develop specific skills for that? Let’s suppose that you would need to mentor a new junior scientist coming in drug discovery. What would you recommend them to be proficient at, in addition to their own very peculiar expertise, for being successful in the future?

René Hoet:

In big companies you have these professional development career programs for scientists, not just on the science base but also on management skills and leading a group. That’s a big thing if you come from academia and you suddenly are responsible for managing a group. You have to learn those things.

Personally, I like teaching young people. I’m also teaching at the university in Maastricht. I try to show them what it takes to develop a new drug, from the target to getting an antibody approved. And very often people come down afterwards to me (these are master students) and say “We would like to know, what is it like to work at the company?” Because those people usually have only academic experience. I say please go ahead, try out, try to get an intern period and work for free to six months, and try to understand if this is what you want to do, or you want to stay in academia. Because that is still a hurdle for people coming from academia – to make that choice early on in their career.

Nicola: What do you think is the role of academia in terms of innovation in the industry? You would consider a key role or what’s your experience there?

René Hoet:

I think academia plays a key role in bringing up new concepts in my view and new ideas. It’s very important that there is a bridge between academia and the industry because many of these ideas stay in academia because they’re not far enough developed so that biotechs or pharma can pick that up. So that I think is a hurdle which is still there in academia. I do think that there are lots of new innovations there. Making the bridge to applications is more difficult. Some of these academia’s try to commercialize it themselves. If they don’t start an official company, I think that’s probably not a good idea. You really need to make a choice there, what you want to do and then fully go for it. And there are people of course that successfully could spin off a company from university and developed it into a company. The successes like Genmab and Argenx are basically started by academic founders.

Nicola: Do you think there are other key elements to transfer the opportunity from academia to industry? Are there some lessons learned from your experience regarding that?

René Hoet:

Early in my career I was working for a company Target Quest, which was founded by Hennie Hoogenboom, one of the pioneers in phage display. He had also an academic group. And he managed to convince the university that he could set up a company next to his academic career. That was a difficult process I can tell you because the university saw that the company was successful and this started a legal fight. That was early in biotech. It’s now probably better arranged in academia. But yes, it is a hurdle to develop technologies from academia to a biotech or pharma environment.

Nicola: Is there a question that you would have wanted me to ask you that I didn’t and you would like to answer?

René Hoet

I ended up in the industry but early on in my career I had to make a choice. I thought I would always be an academic in my whole career early on. Then I got a job in the University of Maastricht, where I was 50% working as an academic and 50% for this company Target Quest. And after a year, I got a permanent job offer at the university. But I was so fascinated by a small biotech company because we started with one-two persons and, within one year, we were 10 people, just financed by CRO work we were doing. I chose the company, and I never regretted that actually. Some people have an understanding of companies that everything is a secret and you cannot share things and that’s a wrong thing in my view. Within the company, there is also a lot of innovation. And so that’s maybe my personal thought of my career. I do like university and academia because I’m also a part-time professor here at the University of Maastricht, but I think developing new products is very fascinating and yes, I really enjoyed that path in my career.

Nicola: I think that’s a great story and I think a lot of people that made the switch from academia to industry can relate to that. Also, in my own personal experience when I was a postdoc, at that moment, for me it was clear that I wanted to change my career path and I wanted to go to industry and start an entrepRenéurship journey. And I clearly remember that moment at that point in time in a similar way in which you are recalling that moment for yourself in which you wanted to experience more of that biotech.  Why do you think that people hesitate to do that? Is just because they think there is not going to be enough research? What do you think is the issue there?

René Hoet: They don’t know so much about it. At least at universities, they are exposed mainly to the university. And therefore, I think it’s also important that people like you and I, for example, also tell a little bit about what happens in industry and motivate people to take those steps.

Nicola: There is life after academia.

René Hoet: Exactly.

Nicola: I can totally relate to that. A lot of great research is actually done in the industry. And of course, there are sometimes different goals, sometimes different timelines, but still, it’s a great place to see and execute on innovation.  So that, I think, is an important message to give to all the younger people that are asking for the same questions.

Nicola: I like often to terminate this kind of conversation with an open question, a bit challenging, to see what people think about the present but also about the future. And this is based on a contrarian question. Peter Thiel often asks that question, which basically says: what do you believe to be an important truth in the field of drug discovery, which is your field, that you believe to be indeed really true, but your peers and others would disagree with you?

René Hoet:

My drive is innovation, new concepts, and new products. In discovery, I usually choose the new steps and the new areas. For example, phenotypic screening, we already talked about this so I don’t want to tell too much about it. But it is an area where you can find innovation. I can tell you, although some of my Bayer colleagues liked it, in a big pharma company this is very difficult to actually get support for. Most of my colleagues didn’t agree with me and said “Let’s do just the targeted approach. We know what we have after a year.”

It’s about risk taking. I think that’s where I differ probably a bit from for most people in the field.

Nicola: We should be all be taking a bit more risk. That’s your final comment?

René Hoet: Yes.

Nicola: René, it has been a pleasure to talk with you and have this conversation. I think we learned a lot and you had a great career. Thanks for joining us. Thanks for joining the show.

René Hoet: Thanks very much for giving me the opportunity. I very much liked the conversation. Thank you.

Nicola: Absolutely.