#91: Marco Schmidt - AI in Drug Discovery and Drug Development: What You Need to Know

What is the Role of Artificial Intelligence in Drug Development and Drug Discovery?

⭐ In this episode, we talk about:

• Deep Mind’s Alphafold
• Investing in Life Science companies
• The State of AI in Drug Development and Drug Discovery
• How does GDPR limit the development of AI in Drug Discovery
• The Human Genome and AI

Today’s speaker is Marco Schmidt, CSO of Biotx.ai
Marco had an exceptional academic career. He won the best PhD of the year award in Chemistry in his home country Germany and then went on to do a Post-Doc at the University of Cambridge.

⭐ EPISODE Links:
Marco Schmidt on Linkedin
Youtube

📖 Quotes:
(24:45) “Eroom's law is the observation that drug discovery is becoming slower and more expensive over time”
(56:05) “We need more data”
(58:13) “Biobanks provide the infrastructure that we really need in the future”
(01:16:39) “Prediction Models of Type 2 Diabetes Have 99% Accuracy”
(01:17:50) “There is no longevity gene. There are shortevity genes.”
(01:53:45) “We are living in fantastic times”

⏰ Timestamps:
(00:00) Kick Off and Intro
(06:15) Deep Mind’s Alphafold
(07:06) Background to Marco Schmidt and Biotx.ai
(12:00) Business Model in AI for Drug Development – from Service Provider to Developer
(13:15) Basic Definition of Drug Discovery and Drug Development for the Listeners of the Episode
(17:20) For Retail Investors who have an eye on investing in public drug development companies: What is a Ligand?
(20:09) Why is Drug Discovery Expensive and Lengthy?
(23:30) Success and Failure Rate in Drug Development
(24:30) What is erooms Law?
(27:25) What is the right place for innovation? Startups or Big Corporations?
(30:00) The Value Chain in the Pharma Industry – From Research to Marketed Products.
(34:00) Investing in Life Science Companies
(50:15) Why is Artificial Intelligence an Attractive Area?
Speaker and Company
(55:00) The State of AI in Drug Development and Drug Discovery – What is Science and What is Still Fiction.
(58:00) Accessibility of Data – Is this the bottleneck in Artificial Intelligence Based Drug Discovery?
(01:10:00) What would change in Drug Development if scientists could access healthy population data?
(01:12:30) Prediction Models for Cancer, Alzheimer, and other diseases
(01:16:00) Understanding the progression of diseases and the role lifestyle plays in AI
(01:19:00) How does GDPR limit the development of AI in Drug Discovery
(01:24:00) Data Curation in AI
(01:28:45) Is Capital a Limiting Factor?
(01:34:26) The Human Genome and AI
(01:40:00) Is it possible to simulate the entire drug development process from lab to market?
(01:46:00) Service Offerings of biotx.ai
(01:51:30) How can AI help select the right patients?
(01:59:00) Synthetic clinical trials
(02:01:00) The Role of Excel in Machine Learning

📚 Book Mentions:

Climate Confident
With a new episode every Wed morning, the Climate Confident podcast is weekly podcast...

Listen on: Apple Podcasts   Spotify

The LSG2G Partners
Experts in Life Science

Disclaimer: This post contains affiliate links. If you make a purchase, I may receive a commission at no extra cost to you.

Transcript

00:00:00:03 - 00:00:17:01

Marco Schmidt

So this is actually, I would like to call it. We are living in fantastic times now because everybody can now do fantastic research together. This computer from home. So because if you need more computational power, then you can use cloud solutions, for example.

00:00:17:10 - 00:00:54:15

Christian Soschner

What is the role of artificial intelligence in drug discovery and drug development? In this episode, we talk about Deepmind's AI four fold investing in life science companies, the state of artificial intelligence in drug discovery and drug development. How does GDP are limits to development of artificial intelligence in drug discovery and the human genome and artificial intelligence. Today's speaker is Marcus Schmidt, chief scientific officer of Biotherapeutics Drugs, a.

00:00:55:02 - 00:01:21:24

Marco Schmidt

Canadian billionaire still. So when they were young, they wanted to be rich. And they're rich. They wanted to be young. I'm coming to your question. So this reminds me to a project you also had. So we looked into the genomes of long living people saw 101, another 20, and so they did check them in the genomes and compared to people who died.

00:01:23:07 - 00:02:01:04

Marco Schmidt

And this there was always the idea that, yes, the longevity gene, but actually what we saw this normal longevity genes are called short deleted genes. So they are really genes that are known for diseases. And if you die early, they are these genes are extremely overrepresented in this in this cohort compared to the long living people. And the other way around is what is courtesy of your prediction because there's still a chance that this is just nonsense.

00:02:01:24 - 00:02:24:06

Marco Schmidt

So you really have to prove it. And to prove it means that you have a prospect of clinical trials. And these trends in 30 years. And it's very expensive to prove It's very expensive and it takes ages to prove it. Yeah, definitely, definitely. Saw, for example, type two diabetes. And we have extremely, extremely good prediction what it's not.

00:02:24:19 - 00:02:57:06

Marco Schmidt

So I think it's 99% accuracy. U.S. and prediction for type two diabetes. The big provide us authentic clinical controls and you not only hear me say they have is they also use a sophisticated, sophisticated algorithm. But in the end, you can also do machine learning was excellent. I know. I know a lot of people and you know, I guess probably it will end up not as the system, but it's possible you can do it.

00:02:58:13 - 00:03:07:06

Marco Schmidt

And believe me, I'm big organizations. Microsoft, excellent student and.

00:03:07:07 - 00:03:47:13

Christian Soschner

My cohort, an exceptional academic career. He won the best Ph.D. of the year award in chemistry in his home country, Germany, and then went on to do a postdoc at the University of Cambridge during his time at Cambridge. Michael received the Marie Curie Fellowship as well as says depending on from the Gates Foundation to do his own research, Michael worked with excellent people like Isabelle would later become vice chancellor, and Tom Blundell, the founders of Aztecs, a fragment based drug discovery company that was acquired for $866 million.

00:03:47:22 - 00:04:26:19

Christian Soschner

His work with Chris and Tom opened his eyes to the possibilities of launching his own company, as well as to the limitation of traditional approaches to drug development. The inability to predict clinical efficacy. Solving this problem became the mission of Biotherapeutics. Today I biotherapeutics Today I combines expert judgment, mechanistic validation and artificial intelligence. The Experts Drug Selection Committee, led by Jake Scannell, picks treatments most likely to go smoothly through regulatory processes, have the market fit and succeed in trials.

00:04:26:22 - 00:05:01:02

Christian Soschner

Heritage treatments are discovered using their unique wide data algorithms which decipher the complex interplay between genes and thereby identify connections between drug targets and diseases that would otherwise remain and see therapeutic stocks. I evaluates data from this and that clinical trial platform. The platform evaluates each potential treatment in an independent mechanistic model. It predicts efficacy and side effects.

00:05:01:14 - 00:05:29:04

Christian Soschner

I hope you enjoy this episode the same way as I did. Marco, it's good to see you. We are now life on the internet, on YouTube, published in and ancillary recording stations to make sure that we don't lose any information of this valuable podcast. It's good to see you today. Yesterday, when appropriate for the episode, I saw a post on LinkedIn from Lex Friedman.

00:05:29:04 - 00:05:59:07

Christian Soschner

He is a podcast from the United States, and he wrote, Maybe I can share my screen and can can show you to post. He congratulated DeepMind for developing Alpha Tensor and he said these are the early steps in inventing new ideas in math and physics that's of often Nobel Prize and Fields medal belief in exciting times. And this morning I pulled up the nature publications.

00:05:59:07 - 00:06:14:15

Christian Soschner

The article he referred to Deepmind's A.I. invents faster algorithms to solve tough maths parcels. Did you hear? People said to me you referenced that DeepMind in in your article on LinkedIn.

00:06:15:06 - 00:06:45:18

Marco Schmidt

Yeah. So that's the company DeepMind. But I mention I fell for it. I support this deep learning of the application to predict the protein structure just based on the sequence. I need to say that I don't have. I haven't read the article, so but it looks for me that this is some other area. Where is DeepMind active? But I only I'm only about five for it now, so.

00:06:46:24 - 00:07:12:03

Christian Soschner

Yeah, but it's pretty exciting development of AI for for this US today. Just articulate and I'm curious to hear more about the what can artificial intelligence store in track development and track discovery. Let's start with the first question. Can you please give me a little bit overview of where you're coming from, what your profession is, and what your company is doing?

00:07:12:23 - 00:07:41:01

Marco Schmidt

Yeah, so my name is Michael. I'm originally a biochemist, so I studied biochemistry in Germany and to being in Berlin. And then I went on to conduct a PhD in fragment based discovery, and this was then followed by a postdoc at the University of Cambridge, mostly funded by the Bill and Melinda Gates Foundation. So I moved more and more into the field of drug development, of drug discovery in my academic career.

00:07:41:01 - 00:08:24:21

Marco Schmidt

And then I worked in Cambridge, and you just said that, Brilliant guys. So I'm Chris Ables, and that's two years ago, unfortunately, and Tom Rondell and these guys are the attacks on us. So they founded the company called Sticks and they took this courage, I think imagine I was nine and then they saw this risk in 2013 to come and what I learned from these two guys especially was the setting this Bill and Melinda Gates Foundation is that the big problem is not to find elegant for drug discovery or drug development project.

00:08:25:09 - 00:08:59:17

Marco Schmidt

It's more problematic to predict the clinical efficacy your silicon. So this is actually the black box. So in the beginning, when you think about the disease, how you want to target the disease and always wanted to use. So you read a lot of literature, you look at in vitro experiments and so on, but these are all small experiments and the big funny experiments, when you go into human and you really want to see that there is an effect or not, this actually, you know, prediction available.

00:08:59:17 - 00:09:30:11

Marco Schmidt

So this is actually a black box. And in 2013, there was in the first publication by the former head of R&D of the US, Merck, so Edwards colleague and he had you proposed the concept of genetic supplied or genetic evidence. So you set the the drug target. So normally you have elegant and elegant brains to really get to a drug target to protect you, mostly protein and modulate the function of the protein.

00:09:31:01 - 00:10:00:15

Marco Schmidt

And then the protein somehow connected in a genome wide association study with the disease. Then you do towards this drug target linkage, genetic evidence. And if you haven't postulated, then it's more likely that the drug target is causal for the disease. So modulation of this would have some effect in the clinical trial. And this was improved in 2015 and a prospective study by people from GlaxoSmithKline.

00:10:01:09 - 00:10:23:10

Marco Schmidt

So they looked at all successful and successful drugs in the last 30 or 40 years and they really could show the tests, this evidence. And then you have genetic evidence and you have genetic supports, and it's more likely, I think it's three times more likely to test the efficacy. And efficacy is shown in the first in a clinical trial as patients.

00:10:25:08 - 00:10:50:24

Marco Schmidt

But the point already is that with just what I call explanatory statistics, so you look at the and you explained the data shed by reject say I the idea was and then I met my co-founder John a computer Lewis So I'm a biochemist he's a computer linguist. Was the idea to use predictive statistics what call called machine learning or artificial intelligence.

00:10:51:12 - 00:11:22:05

Marco Schmidt

So just, you know, that kind of thinking. So we are looking we try to find know the patterns in these genetic and the genetic evidence to use this for prediction as a drug and a drug target combination may be successful in the clinical type for a specific disease or not. Yeah, it was the idea in 2015, 2016, we had then the so-called startup grant.

00:11:22:05 - 00:11:44:10

Marco Schmidt

That's the university of Potsdam. After this, we then joined the course and you have to say we joined Startup Digital Heads Balloon. So Accelerator Program, and then we also founded then biotechs, the idea of a company. And, you know, we were looking more than four years on this and.

00:11:44:11 - 00:11:45:06

Christian Soschner

Congratulations.

00:11:45:18 - 00:11:47:04

Marco Schmidt

On five years. So five years.

00:11:47:19 - 00:11:53:08

Christian Soschner

So you are all what, the best face of a startup. So the first three years are crucial. And then come to big success.

00:11:54:07 - 00:12:07:22

Marco Schmidt

Yeah, actually. So I'll be talking to also from the business model, from the service provider and to know our own products. So we have three products in clinical development and one clinical trial running.

00:12:08:09 - 00:12:12:17

Christian Soschner

So how much money did you raise so far? I'll just curiosity if it's not confidential.

00:12:13:00 - 00:12:21:08

Marco Schmidt

It's it's not confidential. So it's actually not so much the 3 million and the rest of the sum was revenue revenue. But it's.

00:12:21:08 - 00:12:52:21

Christian Soschner

Very efficient. I had two recent podcasts where it was more focusing on entrepreneurship, on investing, and also doing research and development in artificial intelligence. Let's stay in this podcast episode on the artificial intelligence side, and let's start with some basic terms and with some basic definitions. I know from the last year that many retail investors business angels got curious with the successes of CRISPR and Intellia to invest in life science projects.

00:12:53:20 - 00:13:23:01

Christian Soschner

But it's not the most easiest field. I mean, that's not just the first term that you mentioned drug development and drug discovery. When I entered the life science industry in 2006 and it told people that I'm developing drugs, they looked at me and said, What are you doing? Something illegal? Yeah. So 1208 In the times of cancel culture that this podcast episode is taken down, let's start with defining drug development and drug discovery in the context of this episode, what is your definition of drug discovery and drug development?

00:13:23:20 - 00:13:50:10

Marco Schmidt

So first of all, I sort of make it clear so this is not illegal. I know that these two words because in the German speaking system there's often a misunderstanding. So here I would like to focus. Yeah, we are working on medications, so we are working on different kinds of chemical or biological entities that can modulate some stuff in your body in order to treat or cure disease.

00:13:51:21 - 00:14:39:14

Marco Schmidt

Then the next what you asked the drug discovery and development. So normally discovery means that you really are in the early stages of your development. Discovery means and that you you're still think about the disease, but as a mechanism of the disease, how you can treat the disease. And then you start developing, you can try to test the ligands and in vitro models, So in disease models, and that's why they still use discovery thing, because it's like the land is not completely new, it's still unknown and you are the first entering it development a little bit more.

00:14:39:15 - 00:15:04:24

Marco Schmidt

Then you have already proven that you need any drugs and you want to go into the normal way. How to prove it. And you can see, as I call it now, five phases. The first, the so-called preclinical phase that you take your ligand in neutral, it's that it's safe. The next one is the so called phase one clinical trial, and it's still looking at safety.

00:15:05:04 - 00:15:31:19

Marco Schmidt

So you tested that in humans and you're very interested that these humans do not have any severe side effects. So you tested actually enhancing humans to check the safety. And then the clinical phase two becomes, from a development standpoint, much more interesting because this is the first time then you have seen so-called dosage finding. So you want to think about what do you call it to see particular?

00:15:31:19 - 00:15:52:03

Marco Schmidt

No. So what kind of tools that you have to administer, which shows an effect but is not toxic to the person. And here in this, especially in the later phase two clinical trial, because this phase to be this the first term and then you also look for efficacy for the efficacy signal. That's why this is the biggest problem.

00:15:52:14 - 00:16:43:22

Marco Schmidt

So enormous success rate in the phase 2 to 2 to clinical trials, 29%. So it's extremely low. It's extremely low. And then this and clinical phase three, and this is mostly the big cohort that you have to show from a statistical point of view, the test efficacy. So you need a lot of patients and as you control it's it should be randomized placebo controlled perspective to really show that this is an effect and nowadays you have also what you call phase four and phase players after the approval that you your your medication is to under observation by a U.S. development developer and also by the authorities to look whether there are any side effects that

00:16:43:23 - 00:17:03:10

Marco Schmidt

are know and the use of a whole you in the clinic in the daily routine is so probably be some difficulties you haven't haven't seen the testing so this has five phases and five stages you normally so you should subtract development phase and everything before this so-called discovery phase.

00:17:03:23 - 00:17:24:06

Christian Soschner

That's an excellent definition. Thank you very much. Let's also define a second words that you use quite frequently in this episode. Content. People interested in investing in practice, covering abstract development projects. You might also hear in the future. It's elegant in the context of this episode, but it's elegant interactive discovery.

00:17:24:06 - 00:17:59:16

Marco Schmidt

Yeah. So what is elegant? And I would like to say it in a very simple it's actually the tool and it's from a business perspective, your asset. So an elegant is can be a chemical structure or it can be also a biological entity, like an antibody or a peptide. But this is actually something what you have, I call it what you have on your desk and what you have to take.

00:17:59:16 - 00:18:35:21

Marco Schmidt

So this is actually you twin in the ligand. It, by the definition binds to the target to a drug target. So mostly it's a macro molecule. This is deep, mostly with deep pockets, but binding and combine into and by binding this and the change in the function of this macromolecular of the protein, for example, and with this modulation of the function, you you want to induce an effect in the body.

00:18:36:08 - 00:19:04:06

Marco Schmidt

For example, the you famous are a settings for cholesterol. They buy into an enzyme which is responsible for cholesterol synthesis, and they block the active site. So new cholesterol is an animal produced in the liver, for example. So the ligand this is your twin in in in drug development and drug discovery and this but it's also used because I only contract it.

00:19:04:23 - 00:19:27:22

Christian Soschner

I like to elect a business approach that you have. I didn't try this definition. Elegance is your asset. Whenever you hear representation and someone a scientist is talking about elegant, listen closely. He's talking about your assets. So this is this is the most important part of the presentation. I think this is great when we stay in track. Discovery.

00:19:27:22 - 00:19:57:11

Christian Soschner

I think a lot of people now know drug discovery from the last three years. We had all the Biontech and Moderna stories. We've Phase three studies before 2000 patients and also the price point of 3 to 4 billion. And when you look at the length of drug discovery and track development process, I think in the early days of marijuana technology in the sixties and Biontech started somewhere in 2008.

00:19:57:11 - 00:20:14:21

Christian Soschner

So there we are working really hard for over ten years on moving the asset forward to a stage where they can produce a new vaccine. Over the weekend. So this was in the news. Why is this process so expensive, so long and so challenging.

00:20:17:20 - 00:20:46:19

Marco Schmidt

And just a lot of thinking about this And and I don't know that I'm the right person about this because I think I'm still too young, that they have a lot to talk about. So so, yes, other people took my personal view and this is just my personal view set of the day how we do on drug development, drug discovery is still very would be quite reductionistic.

00:20:46:23 - 00:21:13:11

Marco Schmidt

So in science in the last 200 years, we have this mindset of reductionistic science. So what does it mean then? You have a big problem. You separate a big problem into smaller parts, into smaller problems, and then you try to solve some of the problems. And when you have the solution of smaller problems, you think you can put these smaller solutions together and then you have a big solution.

00:21:14:03 - 00:21:40:08

Marco Schmidt

So this is this is very my opinion, very, very successful of this reductionistic approach. So think about your problems. So how you approach this. And it's often that you start separating the big problem into smaller ones and then put it together. And the problem is this living systems, especially with the human body, is that it's more than its parts.

00:21:40:17 - 00:22:11:23

Marco Schmidt

So in this is, in my opinion, the big problem. So you always see that in the we tend to have very simple models in biology because the biology is extremely complex. Our tools are also very limited in still to say it in my opinion. So you have your pipettes, you have your in vitro viruses and so on, and you tend to have very simple models.

00:22:11:23 - 00:22:32:13

Marco Schmidt

And then based on these simple models, you have an idea of how, how, how the ligand, how the drug is behaving in your in your body. And then you can go, for example, from from your cell dish to an animal model, it's obvious or quant. It doesn't work. You know, because the animal is more complicated than I expected.

00:22:33:06 - 00:22:55:22

Marco Schmidt

And then the next step to the human body is also very it's very, very complex. It's very, very difficult to apply this from the very simple model than to the human body, because there's a lot of human bodies, a lot of black box, this lot of things we do not know that we do not understand, but we really want to manipulate the system in a specific direction.

00:22:56:04 - 00:23:14:22

Marco Schmidt

And this is my opinion. This is what makes it so difficult and very risky and so painful because if you do not understand the machine, then it's all in its hole and you want to manipulate it. And then it's extremely difficult.

00:23:14:22 - 00:23:53:03

Christian Soschner

That's absolutely true. The human body is still a mystery, even if just think I read some articles recently about longevity research and the effects of sleep, nutrition, of exercise, life habits. I think this all has an effect on technological systemic changes. Also, the environment for scientists like you in different patient populations, it's absolutely clear when I look at the success rates of drug discovery and early stage drug development, we are talking about 99% failure rates before a project reaches the clinic.

00:23:53:03 - 00:24:15:13

Christian Soschner

So one out of hundred makes it and the other 99 are for publications and to increase the know how interesting knowledge. And when I look into the clinics, I think it's a little bit better. But we are still far from a face saving virus. And remember from a phase one project from ten phase one projects, one reaches to market science start and end, don't you think?

00:24:16:09 - 00:24:36:22

Marco Schmidt

But do you have to differentiate? So big pharma companies are actually not so good an R&D productivity, but biotech's much better. So you see, I don't know whether you heard about is famous abroad it's called be rooms law. So it's a no you haven't heard about.

00:24:36:22 - 00:24:39:24

Christian Soschner

I don't know that it's explain it please tell me more.

00:24:40:00 - 00:25:12:14

Marco Schmidt

So this is actually the complete opposite of Moore's Law. Moore's Law is when you look at substance abuse, that's you as they become better and better over the time and becomes law as the the that are in productivity is decreasing over the time really. And this is the this is your the overall idea of Ohm's law but how they describe R&D, productivity and big pharma.

00:25:13:12 - 00:25:44:07

Marco Schmidt

But what we have seen since 2010 is that they call it breaking your Ohm's Law, that the productivity is now increasing also in pharma. This very interesting because we see now that just a lot of change in decision making in pharma and biotech based on novel technologies. And the also as a business strategist we see that makes these projects more successful.

00:25:44:07 - 00:25:46:03

Marco Schmidt

In the past.

00:25:46:03 - 00:26:12:16

Christian Soschner

That's, that's amazing. I didn't look at it from this angle. It just looked at it from the organizational point of view and my personal explanation why R&D doesn't work so well in big pharma companies is that by definition a company is an organization. They are more than one person works hard on accomplishing just one goal for a defined set of customers.

00:26:12:16 - 00:26:44:01

Christian Soschner

Just think about April, for example, for most of the time in this century was the one product company that trusted with Apple, iPhones, smartphones and the entire organization is set up to just accomplish this longer. Push one smartphone to the population. And every time when something new is discovered, it means that this new thing that might be beneficial in the future is simply outside of the mission of the company.

00:26:44:13 - 00:27:15:21

Christian Soschner

So it means doing research and doing R&D inside The company can't work by definition because it doesn't really help the current goal in the current mission. So it might help in the future, but counteracts the processes that are set up in the company. And my personal opinion of I was when you find something, you put it in a startup outside of the organization, set up a project team, lets them better connect, and when they are successful, then you can re-integrate it because then it fits in that into the mission.

00:27:17:04 - 00:27:29:01

Marco Schmidt

It's a little bit to innovate, just to deliver a dilemma you're talking about. I guess so that for a big company, what could become graduation and new technology is almost a kind of a threat to the existing business model.

00:27:29:22 - 00:27:32:12

Christian Soschner

That you could call it that way. You could call it that.

00:27:32:24 - 00:28:08:22

Marco Schmidt

You know, I don't know. I am, in my opinion, when I look at my company, especially R&D facilities, in my opinion, it's something even the job to do to to prove new technologies. And people are always complaining about some of it, in my opinion. It's tough. And in the past they had too much focus on this reductionistic towards what I had seen before.

00:28:09:14 - 00:28:50:02

Marco Schmidt

But you can say this the Human Genome Project in 2000, I guess the mid 2000, you see more and more that they are using genomics omics technologies more and more, and they are also going more into very specific disease areas. So they break with the general idea of the big blockbuster. So where you have 1 million patients and then you want to deliver a drug for 1 million patients, then I call it the, you know, the things you have to do it.

00:28:50:02 - 00:29:18:15

Marco Schmidt

It's it's very tough. You know, you have to conduct a big clinical trial. You really have to show the to entrance book and so on. It has to be safe in a safe in a in a broad population. And since 15 years, I would like to say you see that there are more going costs. I call it diseases to smaller patient groups, more targeted where they have more omics evidence that this is working.

00:29:18:15 - 00:29:51:00

Marco Schmidt

And here you see that there's this more successful. So of course the biotech is always faster moving. But for example when you see rice, Genentech, for example, they, they they, they try to set it up to have more startup oriented organizations, Genentech. And you have to rush this, the traditional pharma. And of course, say it's in distribution units you need and you want to market the drug.

00:29:51:11 - 00:30:08:23

Christian Soschner

Of Genentech is a good point. I'm currently reading the book of Sebastian MALLABY about he was talking about the power law and the development of venture capital in Silicon Valley and Canada. And it's a big part of the book. I mean, you mentioned it a think I like I said, the term networked economy that exists since the nineties.

00:30:09:09 - 00:30:44:07

Christian Soschner

I think trying to put everything in a big company. So from the beginning of the research and development up to market in your product is just too diverse. Mission You mentioned marketing and sales. Bringing a product to like the Mariner epic seems to a billion or 2 billion of people in different jurisdictions. It's hard work and I would rather focus a big organization on this goal to distribute, define it as product, then trying to push everything in in the entire value chain.

00:30:44:07 - 00:31:12:13

Christian Soschner

So really like the set up that we have currently, you have big pharma that are proficient in marketing and doing the final battle with the regulators and then the early stage biotechs that solve the development problems that are not really suitable for research organizations due to a different mindset and also not yet ready for big Pharma. And then you have the old ground of pure science and research where I think the research organizations are the best part to have it.

00:31:12:13 - 00:31:20:16

Christian Soschner

So I really like this major chain approach of research. You have the market experts and in between there are biotechs and venture capital.

00:31:20:16 - 00:31:48:20

Marco Schmidt

But also a problem for the biotech company because of the value creating step or value make some money as it is. And pharma. So there is this famous iGEM. When you have invested $1 in the biotech stock and $1 in the pharma stock, then the biotech, your biotech investment is another 16 US dollar. You have invested in pharma, it's now 800 US dollar.

00:31:48:20 - 00:32:15:12

Marco Schmidt

GROSS So you see and this is actually one thing that the biotech people have to think about, how they can capture the value still because all the value is created in the last step and all the money is spent in the last step. So it's very difficult to keep, you know, how you can keep up as a success in the next stage is the idea in the early stage.

00:32:16:11 - 00:32:19:03

Christian Soschner

The negotiation talent in between.

00:32:19:10 - 00:32:20:09

Marco Schmidt

The and the user.

00:32:21:24 - 00:32:23:05

Christian Soschner

Interrupt me just just go ahead.

00:32:23:17 - 00:33:19:04

Marco Schmidt

Know what I mean as you see. And then we can also apply this to the idea of discovery development, trace industries. So the startups in this area, we saw they all started as a service provider last spring, for example. And then you move over the time into a clinical stage company because you see the real value is at the end and there was again a two weeks ago there was business article with the same guy, it was by the same person who coined such a quantum name, Iran's logic scanner, and he published a game theory paper about why is a biotech companies, especially the platform companies, are now going into move into a clinical biotechnology

00:33:19:04 - 00:33:49:11

Marco Schmidt

company because of this reason money this little stage. And this just underlines it's because in the beginning when you have some an innovation, people do not tend to invest in the early innovation because they do not see the money back or it's just suddenly very interesting from a innovation point of view, how you can get loans or the beginning, the really new stuff can come.

00:33:49:11 - 00:33:57:15

Marco Schmidt

We need to patients when there's no incentives anymore for investors to invest in early stage projects.

00:33:57:15 - 00:34:18:07

Christian Soschner

I can give you my my perception of this area. It's just my definition. I think that's come in in Europe, you have a general problem. There's not enough capital on the market for the innovation that we have because it is it's in a company of the availability of venture capital in Europe with the United States and also with China.

00:34:18:07 - 00:34:43:06

Christian Soschner

They are the other two regions, China, the United States invest 5 to 10 times more per year in their companies and in their research, and especially in development. When you look at Europe, I think the dollars of the maturity of the euros go into early stage research, not development. So there is a lot of grant funding, a lot of philanthropic approach.

00:34:43:06 - 00:35:10:03

Christian Soschner

As you also mentioned, Bill and Melinda Gates, for example, the foundation is our philanthropic approaches. So whether they invest in companies or in the research organizations, not to develop a track up to the market, but to just produce new IP, you know, and not finalize projects when you look at the later stages. So there is more money. I mean, to risk, there's less risk and the stories are just bigger.

00:35:10:03 - 00:35:37:08

Christian Soschner

So you need to allocate more money. When you look at this stage, you see in between to are investors, but the risk profile is different. So you mentioned Merck, Pfizer, GlaxoSmithKline's. So these are the big pharma companies. When they put the euro in these companies, the risk of them going extinct is almost zero. It exists. We can have an economic breakdown or something really terrible can happen in the world.

00:35:37:14 - 00:36:03:16

Christian Soschner

And then these companies and they exist anymore. But usually they are set up to survive. When you look at early stage investments, for example, into startups and know that every year I invest in a startup company has a chance of success when we are in pre-clinical stage of maximum 5 to 10% of 1 to 3% rate. It depends on the technologies, but it's not high up to 10%.

00:36:03:24 - 00:36:31:14

Christian Soschner

If I bet all my money on one company, it's a high risk of failure. When venture capitalist spreads the money on a minimum 15 companies, they can narrow down the risk almost to zero. And usually when I look at the portfolios of venture capitalists, one or two companies repay the entire fund and leave enough profits to motivate their limited partners to reinvest in a second fund.

00:36:31:14 - 00:36:42:14

Christian Soschner

So this model works really well. But the problem only arises when people bet their entire capital on one company, especially interactive discovery.

00:36:43:02 - 00:37:12:16

Marco Schmidt

Yeah, that's huge problem, especially in Europe. As you mentioned that the size of the venture capital of France here in Europe, in my opinion, to small. So when you look at the bigger firms in the Silicon Valley, so you you see all this a very, very excellent quality, very excellent expertise or expert department, but they always have. And this is what you complete.

00:37:12:24 - 00:37:40:07

Marco Schmidt

You are very here in Europe, you don't want to see. It's it's not there, but it's very limited because especially for Crispr-Cas, for example, let's discuss our expenses. And the reason why I should join with you is for European, probably not completely clear. So we stay in it's university or at a pharma company about a biotech, for example.

00:37:41:00 - 00:38:10:15

Marco Schmidt

But we see a lot of people, especially in the genetic state, moving towards jobs in venture capital. Now, and you can then judge that investment. But this is very, very, very expensive to hire all these guys to have the expertise in your in your resume. And it only makes sense then you have to be very, very, very busy because then you have this two or 22% is the management fee.

00:38:11:01 - 00:38:37:02

Marco Schmidt

But it's it's it's for example, billions, I don't know, 100, 200 billions of volume and it's in 2%. And then you can afford these people and then you can make also better decisions. This is also what I see that you have some kind of an advance advantage in your expertise in the and we see then you have a bigger we see fund and.

00:38:37:12 - 00:39:02:08

Christian Soschner

I completely agree to what you say. I think the the two main points for me are whenever I help scientists move the company forward, the success story is straightforward. You can find the company in Europe, you can get a lot of grants and early stage support. So the minute a scientist has a breakthrough, the invention in a research organization, there are sufficient funds for the early stages.

00:39:02:22 - 00:39:25:10

Christian Soschner

But the genius and you mentioned CRISPR, for example, it's and I think the story of röttgen over contributor RepuTex is an excellent example for how they processed and continuous after the foundation. So as far as I researched it, wrote to Novak and Emmanuelle Charpentier, met in Vienna and then decided to create the company in Switzerland and got early stage funding obviously, otherwise there wouldn't be a company.

00:39:25:23 - 00:39:48:17

Christian Soschner

Then the next step was clearly to pass them so to the United States to fund the development in the later stages. And this I think, is the big downside of Europe. We are caught in early stages, but our limited funds at a later stage, you and what you sets the size of the fund, anything below €10 million in advance doesn't make any sense.

00:39:48:17 - 00:40:11:24

Christian Soschner

On one hand, you need the experts in the fund. On the other hand, you need to have enough capital to allocate in a big enough portfolio to have two winners, at least in a pool of 10 to 20 companies. And for mid-stage funds, I think anything below 50 million. So mid-stage preclinical clinical phase one, it's a little bit late.

00:40:11:24 - 00:40:21:12

Christian Soschner

Early stage up to mid stages is the minimum size of the fund, 50 million and for later stage funds, 100 to 200 million. So it's this that they're doing is.

00:40:22:04 - 00:40:26:19

Marco Schmidt

So so you're means of complete the total volume or just per investment.

00:40:27:06 - 00:40:29:13

Christian Soschner

To total volume to fund size. The fund for.

00:40:29:23 - 00:41:01:11

Marco Schmidt

The I always is 50 I think and still it's due to less money, I think even 100 then you have 100 you can do probably 1015 investments. So it's all there so that you have to keep some dry powder. For example, just the recession, then it's difficult to find the next funding round and you have to put more money into your investments, otherwise you die.

00:41:02:00 - 00:41:32:18

Marco Schmidt

So I think in the this 10 to 15 million in the biotechnology space in tech, I think it's different because you are you are faster to some kind of a service or product, but you can show revenue ads and also that you also find these investors, they are more expensive is e-commerce ideas they can raise and also bigger rounds and so on.

00:41:33:12 - 00:41:43:09

Marco Schmidt

But here in biotechnology, ten 50000000 minutes for Hunt in my opinion like suicide anymore.

00:41:43:20 - 00:42:14:01

Christian Soschner

It's I completely agree it's the minimum to to start the first fund so when someone wants to start their first fund it makes sense to start to get the first fund off the ground. It's even harder to raise money for a fund for a first time fund when the team comes together for the first time. So starting before 10 billion round to prove it works and it's the diversity still from expense, but then 5055 0 million is the minimum is the minimum for the next fund to play endlessly in this game.

00:42:15:03 - 00:42:36:06

Marco Schmidt

And it's actually extremely hard to raise the funds and you're your startup when you see it's extremely hard because you are investors and this is a frontier one to see success in. How you want to prove success is that you for this, you need money to invest. It's, it's, it's tough. It's really tough.

00:42:36:15 - 00:43:09:07

Christian Soschner

And it's still an open job for regulators. If any politician is watching, we need to motivate the institutional investors, banks, insurances and also pension funds. It's a huge difference in the United States. Pension funds, for example, they are not allowed to provide venture capital. And when they would allocate also in Europe, only 1 to 3% of their capital under management into venture capital, it would be paradise for drug development here in Europe.

00:43:09:07 - 00:43:35:12

Marco Schmidt

But that is the general problem I see in Europe that I call it financial engineering, that this is not so common in Europe. I know. So when I started by then we started a company in Germany. It was extremely common that when you reinvent the business into, for example, this business, I really wanted to have shares in your company.

00:43:35:23 - 00:44:05:16

Marco Schmidt

So you have to go to the notary. So and usually you tickets in this time 25,000 or 15, 50000 to 25000, but then you spend 1000 or 2000 just for the for the last and the notary. And then four years ago or five years ago, then you had this convertible loan. So I don't know whether you wrote about the Y Combinator convertible loans or it's actually a template you can download from the Y Combinator.

00:44:06:00 - 00:44:34:23

Marco Schmidt

And this is this has huge impact the cost in the early beginning. And also it's not really clear what is the value of which is a valuation of your company. You just test is colloquially known and then this question can be answered in the future. Then you may make a real investment round the numbers on the table and you can say to your own investors for sure, why this and this discount what you have agreed in your convertible notes.

00:44:35:04 - 00:44:40:15

Marco Schmidt

So I'm a huge fan of. This contract in the North and these are from the US originally so.

00:44:41:01 - 00:45:16:17

Christian Soschner

Completely agreement which allows for a Y Combinator funded safe model. I started fundraising in 2006 in the industry and there was no standard on the market how to get capital into a company. And sometimes I got offers of €100,000 for 90% of the company, which doesn't make any sense. And Y Combinator just issued. I think it's I don't have the latest version in my but I think it's 7% $425,000 in a in a convertible loan in a safe structure with it is really great.

00:45:16:22 - 00:45:26:01

Christian Soschner

It sets it sets a great standard for early stage investments for the founders and for the for those organizations. Who wants to bet in early tickets?

00:45:26:20 - 00:45:57:08

Marco Schmidt

Yeah. So it's a little bit about valuation. And so an investment in, say it's very interesting. So for example, last couple of weeks I heard from from a friend so he wants to raise 2 million and then we are offered some percentage of three of his company. And then you also also then talk to some U.S. investors and U.S. investors, they send it and say, you're not ambitious enough.

00:45:57:21 - 00:46:32:13

Christian Soschner

Yeah, it's a different world. It's a different world over there. See Citroen's Digital. The two companies ahead on the podcast. First venture in Silicon Valley from an effects ventures, the Wanted seven plus plus eight. It's a think 50,000,015 one one $5 million. He's located in Israel but has a provision network in in Palo Alto and data company cement stage trust service type of bio.

00:46:33:00 - 00:46:37:08

Christian Soschner

If I remember to think it was also 7 million seed round seed capital seed.

00:46:37:08 - 00:46:37:24

Marco Schmidt

Capital.

00:46:38:10 - 00:47:05:19

Christian Soschner

Bracket development, artificial intelligence. When I look at Europe, when we talk about seed capital, we talk about 2 to 300000 equity leveraged with 1 to 2 million public funding. So we are about 50% of the size that the seed rounds are in the United States. But in the United States, mostly it's venture capital, early stage, venture capital app.

00:47:05:24 - 00:47:30:03

Marco Schmidt

But I really have to say that and keep it in mind, the U.S. is more expensive. So I'm pretty sure if you have two or 3 million rule, you can go much more far than when you have 7 million U.S. dollar. In Boston, for example, keep it in mind because salaries are extremely high in the Bay Area and in Boston.

00:47:30:21 - 00:47:49:05

Marco Schmidt

So and also it's incredibly expensive in the U.S. So you see now some movements of studios coming to Europe because they these opportunities to development is much, much cheaper here in Europe.

00:47:49:14 - 00:47:53:13

Christian Soschner

Especially with the securities deficits. Also an additional upside.

00:47:53:20 - 00:48:08:10

Marco Schmidt

Yeah. Yeah. But this is when you hear about the salaries in in Boston, for example, for specialists, it's it's incredible. It's you know.

00:48:08:24 - 00:48:37:03

Christian Soschner

I see a lot of collaboration potential. I completely agree. I mean, bringing the early stage development to Europe. It's it's cheap. There's a lot of expertise. There's a lot of public funds. There must be some fee, someone who motivates the US investors because you should get there and start to get some tax breaks and tax benefits when they invest in US companies, there is not so much, much incentive to bring us capital to Europe, But from a business perspective, I completely agree.

00:48:37:03 - 00:49:06:12

Christian Soschner

I mean, if they would like to or if they would take the risk to invest in Europe, they could easily find an environment with less export where they could achieve more for each dollar due to which Europe and lower overall expenses in the early stages. Instead in the late stages then to have two hands on a company and can move, then the assets and the team to Silicon Valley or to Boston and completely later stage developments.

00:49:06:12 - 00:49:33:10

Christian Soschner

They are where there is more of an availability of capital. I mean, CRISPR Therapeutics for me is always the best example of how you can move breakthrough technology forward. You let's let's let's go back to artificial intelligence. I started getting interested in digital solutions in 2000, 13, 14, 15, but it was really challenging to talk with investors back in the day.

00:49:35:00 - 00:50:06:24

Christian Soschner

The digital investors, as you mentioned before, were so really interested in courage to market products. Drug discovery is nothing new to that. And the usual drug discovery, drug development this is the still in their environment that they invested in molecules, that there could be problems and that can become the next game changing technology. Artificial intelligence was in between, so it was not digital enough and drug discovery and it wasn't abstract discovery enough.

00:50:06:24 - 00:50:22:13

Christian Soschner

Then, sadly, five, six years later, there is a big hype on the market. So I think every every company that I get in touch with now presents an artificial intelligence solution. But it is hype come from in the last years.

00:50:22:13 - 00:50:50:17

Marco Schmidt

The hype comes, in my opinion, that there's a lot of pressure in pharma R&D, so I used it before as about your Ohm's Law and it's just a little bit like a Democrat in the wind industry. And I know that then you are a former executive and you meet your investors. Investors are basically your own your productivity as a lawyer, how you can, how you can improve it, how you can improve it.

00:50:51:00 - 00:51:17:01

Marco Schmidt

And they ask every time how you can improve it. And it turns out that artificial intelligence, listen, it is a bit like an kind of an escape route and so they say we do artificial intelligence so we know it can do all this on scale because artificial intelligence is and is just a synonym for investors, that you have a scalable product, that your product is scalable, that you service as a scalable.

00:51:17:14 - 00:51:53:19

Marco Schmidt

And this is now also adapted to the industry. And you're completely right. So I see in talks now every company is doing no artificial intelligence, so they have done the same thing before. There is no artificial intelligence. So this example is molecular modeling, for example. So I had to suspect some biology background because of my postdoc. We used a lot of molecular modeling and in those days it was quite molecular modeling.

00:51:53:19 - 00:52:34:21

Marco Schmidt

And nowadays it's called deep learning. I need the techniques in my opinions. I was mostly most like little bit the same. Now, of course, the algorithm, I'm much more advanced for sure, but the data sets and this a assess my opinion. For example, when you look at Alphabet's Google, I fought for it. So for every hour I call it machine learning approach, you meet a very good data science that you can learn from and you kinds of patterns and I for for success relies, in my opinion, more on the extreme.

00:52:34:22 - 00:53:09:02

Marco Schmidt

You get created datasets provided by an army of scientists all over the world. So there are thousands of structure biologists. They sourced crystal structure to last 30, 40 years, and they made them all public. So they are publicly available. So you can go on to the protein protein database. It's it's, it's, it's, it's a website from the Internet and you can download 200,000 structures from there.

00:53:09:18 - 00:53:45:22

Marco Schmidt

And these the security datasets they they had to use. So and I would like to use this more than the algorithm that DeepMind developed because the curated dataset is often a big bottleneck in the development. And it's, it's that's one point that more and more datasets are becoming available when you look at the genomics area, we see a lot of good datasets now becoming available from the departments.

00:53:45:22 - 00:54:03:09

Marco Schmidt

But from the UK Biobank or Genomics, England Genomics, England for example, and these very I call it excellent curated datasets of, the basis knowledge who applies is I call it sophisticated or novel predictive algorithms.

00:54:03:18 - 00:54:46:22

Christian Soschner

It's a great point that you bring up the different facets of artificial intelligence. When I look at the topic from the science fiction point of view, I always thought, wouldn't it be really nice that when a scientist has a new idea that we have a machine, that you can fit the idea into it? And the machine simulates gentle process from the first experiments up to continue phase three trials and based on this data, you can then get regulatory approval at the end of the day of a fully automated manufacturing unit.

00:54:47:06 - 00:55:07:17

Christian Soschner

The truck just comes out and is ready to use for the patients. So this would be a dream, but so that you don't need this lengthy process in between, that you have supercomputers to calculate everything. What's the reality? What are the different facets currently of artificial intelligence and machine learning and drug development that are really reality.

00:55:08:10 - 00:55:32:22

Marco Schmidt

To achieve what you describe? I don't think it's so much science fiction in my ability. So no, no. For example, when I look back at my postdoc and we had a very, very difficult project in structure to solve, but we had selected other structures, very variable, this different confirmation. And what I have learned is, is when you have a good database.

00:55:33:00 - 00:56:06:03

Marco Schmidt

So we had plenty of structures available as a data than molecular modeling makes more sense in what you describe, in my opinion is not the problem of to have a supercomputer. I think it's it's more a problem of having the data. And then we come to a point where we have so much data available. I how to do just simulation between experimental simulated clinical development and so on.

00:56:06:08 - 00:56:10:17

Marco Schmidt

Then I think it's not so interesting what you describe here. How can be how convenient.

00:56:10:17 - 00:56:12:08

Christian Soschner

So how can we get the data?

00:56:12:13 - 00:56:46:02

Marco Schmidt

Yeah, we need we need more data, we need more. And, and very important is that that's created and we have to bring these kind of data to the public resource. I'm and I'm always an advocate to bring data to that to the to do it to the public. What I see here, you have to make sure that in the early beginnings when students do the work and enough to everything controlled under specific conditions, that makes it reproducible for others.

00:56:46:17 - 00:57:19:03

Marco Schmidt

And then you didn't have all these kind of resource and features into your measurement machine. Then it makes sense of the other problem. What you have stirred you in academic is that academic research, in my opinion, is due to biased because academic system works. It's an unequal incentives. It's been supported by as a scientists. So it's more clever to bet on topics that I know that other scientists are also interested in, because then they quote my work.

00:57:19:03 - 00:57:45:03

Marco Schmidt

So that's why we have these courses, clubs in specific areas. And it seems is this is a problem in university and research organizations to go to from these clusters of high quotations, of high citations, more and the other based. And then what you describe it's in my opinion is possible.

00:57:45:17 - 00:58:10:11

Christian Soschner

They are discussed this are good and sometimes challenging at the same time. So since I'm not in academia, I can't remember that. But they found the second point, which is more in my area of expertise, quite interesting. They do understand it, right? Accessibility of data is it's a huge problem. So that's data that's produced in laboratories at research organizations and also in companies.

00:58:10:11 - 00:58:26:19

Christian Soschner

You noticed which companies is not really accessible so that you don't get access to the data to have sufficient data base where you can start training your artificial intelligence on is it's the right understanding or telling me something.

00:58:26:19 - 00:59:10:07

Marco Schmidt

So what you see is that a lot of data which is produced by colleagues in just smaller units, which can be just a research lab, the university and so on, they're very free and what they do and how they conduct their research. So that's the reason why it comes that when you want to repeat the experiment in your lab, then you cannot you can come to the same results because I don't know, there are some other conditions that are actually not described in the protocol, but because for them, it's I don't know, probably the air conditioners are based on 20 degrees and your condition is constantly on 25 degrees.

00:59:10:07 - 00:59:40:21

Marco Schmidt

And this has to be same, except I don't know, but in the earlier in the earlier point, I uses a lot of bias in the research and it's very difficult to be destroyed, bias in the research in the early phase and to make this research and also available because most of them most of this resides, in my opinion, unsuccessful and unsuccessful resides.

00:59:40:21 - 01:00:02:20

Marco Schmidt

I'm not interested to not a scientist to publish unsuccessful results. So there's no incentives for him. So you will not see a citation for this work. But he I guess it's a huge, huge potential. In my opinion. The machine can learn what can be, could be, can do and what we cannot do.

01:00:03:09 - 01:00:29:19

Christian Soschner

Now, the stupid question can go some companies I mean when when I think when after raising funds from venture capitalist sometimes the initial hypothesis hypothesis doesn't simply work and the companies basically shuts down what happens afterwards. If the data is quite simple, it's it stays in a corporate shed. So no scientist has any incentive to publish the data in a publication.

01:00:30:07 - 01:00:49:23

Christian Soschner

And also, devices don't have an incentive to open the data room to public. So it just sticks around in a in a corporate shell very often, in my opinion. Well, sometimes 20. I'd like you to just merge it with another portfolio company. Then it's open for this, that, but for the company. But a lot of companies just don't move forward.

01:00:50:08 - 01:01:04:10

Marco Schmidt

No exits and is added as completely missing. So it takes ages to create the data. This is also one point and nobody is interested in in this time in something that they really don't know whether they a reward for them.

01:01:04:10 - 01:01:11:15

Christian Soschner

Do you have any ideas? Are there any ideas on the market that you are aware of how to solve that problem?

01:01:11:15 - 01:02:05:22

Marco Schmidt

Actually, yeah. You're talking about early research and actually no idea. What I see is the biobank data. For example, there used to be cohorts here. You thousands of participants, for example, and they are really, really very courageous to data. And this has been released in the trade and what you see to see Biobank data and also this our customers who have seen all this say I never experienced some of this from the academic partner or from the academic from the for example, we can test use in vitro data in the population data and then some time it turned out to be true that was not done properly or it's actually wrong and we cannot see

01:02:05:22 - 01:02:38:03

Marco Schmidt

it in the population data, but we see in the population data also the signal, which is what was already postulated. And then it's the great and you can go on. So to answer your question, I think we do not have the resources to clean up all the early stuff, but we can use all the data we have from the later stage, from population based, from genetics, from us and so on.

01:02:38:16 - 01:02:43:11

Marco Schmidt

And she we can prove at least the assumption that from the beginning they were able to.

01:02:45:04 - 01:02:58:08

Christian Soschner

Know it was really great to get access to all the data out to you. So. So basically you need a super creator who gets all the data and structured it in a way that it's usable for data stage.

01:02:58:08 - 01:03:24:00

Marco Schmidt

Yeah, but you see now that it's a biobank. So I share all this to the UK Biobank because these guys, it's just really fantastic what they do. And I'm a little bit disappointed that we do not have a team Continental record, especially in Germany or in the German speaking countries, because I see that this is a kind of an infrastructure we really need for the future.

01:03:24:00 - 01:03:53:22

Marco Schmidt

So people stick too much on the research and on the data. But then when we look back to the pandemic, there were some so called British variant of the British mutation. And this actually based on the fact that only the British had the capacity to sequence enough COVID. The tests, and this is just based on that they had the infrastructure for the UK Biobank, so UK Biobank sequenced 500,000 people.

01:03:53:22 - 01:04:17:13

Marco Schmidt

So when you want to sequence, I find it tells people you need two machines to sequence. And so then the pandemic, then it's for you really to sequence us as of COVID tests to see whether this a new variant or not. And in my opinion, they the UK is extremely benefiting now from the Infrastructure UK Biobank in two ways.

01:04:17:13 - 01:04:44:04

Marco Schmidt

They have now, I call it a proper industry sequencing, gene sequencing, but they also have not a lot of data companies strong data companies in the area and want more of this time you will see the TAB takes a data and the products out of the data are based on the UK Biobank or Genomics England.

01:04:45:00 - 01:05:09:10

Christian Soschner

I didn't I was not aware that if every country uses the data properly that I mean we also in Austria had a lot of tests, a lot of testing. It was free for the population, which is very good. But so we really don't have to necessarily infrastructure to interpret the data after what and to use the data afterwards in in research like the example you brought up in from the United Kingdom.

01:05:09:10 - 01:05:10:10

Marco Schmidt

You mean as Couvert.

01:05:10:22 - 01:05:12:13

Christian Soschner

Discovered, for example, I mean it's just an.

01:05:13:06 - 01:05:44:23

Marco Schmidt

Recorded. So is the British a much more closer to the pandemic tend to be that in the time so you have your participants in the UK Biobank and use that out to complement each of the participants. And then you and you know already that they have been infected, recorded or not, and then they conducted antibody tests, for example, to see how many people were in the whole UK Biobank have antibody and then you have an idea how many people have been already infected by COVID 19 in the UK or not.

01:05:45:14 - 01:06:13:16

Marco Schmidt

Then of course, when they come the are picking the participants, all the participants of supervising come to the physicians and then say, I'm feeling sick. And then at this time then is also going to the UK Biobank facility and then they can see we just only do a PCR positive or negative, but they haven't really tested or sequenced what kind of variant you have here.

01:06:13:17 - 01:06:46:05

Marco Schmidt

So there we are, much more closer to the pandemic and then you're closer to the pandemic. Then it's much more easy to to say, okay, it's worse to shut down here, for example, or what kind of medication is necessary. So I see the I see here because advantage in Germany that you do not have this biobank infrastructure as the UK has.

01:06:47:01 - 01:07:20:23

Christian Soschner

Never been collecting. You name your first post about artificial intelligence in practice, current development, why it the way we do science does not work for track development. And this seems to be one problem. I mean, I personally would not care if anybody uses my my data for drug discovery purposes. So for example, whenever I'm tested or genetic testing or also I mean I have everybody chance that the tracks every time when around adaptive also Apple is investing a lot of capital into improving data readouts from human beings.

01:07:21:12 - 01:07:39:06

Christian Soschner

So if I understood your and your perfect world would be for moving track development forward that all of this data that comes from human beings is structured into a one huge database, then can scientists use for their predictions The simplified right understanding?

01:07:39:23 - 01:08:25:23

Marco Schmidt

Yeah, it's simplified, but I would like to say that when you look at the you could buy it and you could buy tests. And so I think people are overrepresented. I would really point to pointed out here and this is important. So if you had seen that there's actually no need to monitor you, but if you're sick and there's no treatment available for you, then it's good that you do need your data to the organization and to the infrastructure that the research facilities and also biotech companies can use the data to test in city core ideas in a very, very early stage.

01:08:26:18 - 01:08:48:15

Marco Schmidt

Because and this really would I would like just to mention so when this data is available and you test your hypothesis in a very early stage, this is just been so much more towards people cannot think about it to so it's so important. It's a very earliest time in your development to have an idea whether that spoke or not.

01:08:49:08 - 01:09:17:03

Marco Schmidt

And if you can prove it then in the data are then you have also more idea about the end points for a clinical trial. It's a much easier to set up a clinical trial and to find a cure for some kind of treatment. And it's so cheap compared to the standard way so that even you have a rare disease and probably only ten or 20 guys in the world have this disease.

01:09:17:19 - 01:09:42:21

Marco Schmidt

It makes it so cheap that there's still a realistic chance that this clinical can be funded because then points are so clear and they were so good that you can go on this. This, for example, a publicly funded clinical trial to get the cure, I guess the treatment for the disease.

01:09:42:21 - 01:10:06:21

Christian Soschner

I couldn't agree more. There is an interesting point in what you said to said that we only have to take the data from sick people. Wouldn't it be nice to trace it differently? The question what would change if we had access to all data, not only the sick people but also the health of population? Would that be an improvement to track development processes?

01:10:06:21 - 01:10:36:03

Marco Schmidt

So this is a good question, and I would like to call it this way. No, no, no. Nor so because dimension is it's incredible. So for example, we defined 200 million positions, so it's so called single nucleotide polymorphisms. So 200 million positions. My GM differs from your genome. So in all it's 3.5 billion DNA letters, but only in 200 million positions.

01:10:36:03 - 01:11:11:05

Marco Schmidt

We differ. So then we look for interaction in the data sometimes as so that you need to gene variants to have some disease. So it's in 200 million, which applied by 2 million. And then you come to a to a space that's just so big so as not enough patient on this planet. So these 200 million which applied by 200 million, means that you have to sequence 5 million times human population.

01:11:11:17 - 01:11:46:06

Marco Schmidt

Okay. So to get to see the statistic power, so on, I always say and this isn't not cynical so it's, it's, it's, it's just only a very scientific or very rational conclusion is that we do not have enough sick people to conduct proper statistics, massive data on our because the dimension you cannot imagine, because if you have the or 5 million times the you in population you need to cover to understand from a statistical point of view the human genome.

01:11:46:21 - 01:12:02:03

Marco Schmidt

So this is you rich, and I'm only talking about a two way infection. You're talking about just three rate. Actually, it would be 200 million multiplied by 200, which you that 200 million. So it's large. It's it's unbelievable for. Yeah.

01:12:02:10 - 01:12:07:11

Christian Soschner

So I was, I was packing my, my my doctors with one question.

01:12:07:11 - 01:12:07:18

Marco Schmidt

Yeah.

01:12:08:06 - 01:12:34:14

Christian Soschner

When I had my unit checkup. You should have got the answer. You are healthy so come back sick which is a good thing. Yeah, but then I always asked my favorite question. I heard that people get cancer when they are in their forties or fifties. What can it do now? So 20, 30 years ago when I was in my twenties and I'm not 48, what can I do now to avoid getting cancer?

01:12:35:12 - 01:12:41:19

Christian Soschner

So and so that's was no answer. And I was hoping that artificial intelligence could could solve this problem.

01:12:41:19 - 01:13:19:04

Marco Schmidt

But yeah, so some years ago we had a project, so we developed prediction models for breast cancer and for type two diabetes and also for some of US disease. So one of our first project was to develop a prediction model heights in this disease and this is famous gene output. Before it's when you have it, then it's more likely that a some age you will have some disease.

01:13:19:04 - 01:13:44:05

Marco Schmidt

But this can be with a 60, but it can be also this 90. So it's it's it's difficult. So and what's a very nice excellent for a good example for would be quite explanatory statistics because it is extremely overrepresented and I say my patients but it has actually no value for prediction so because it's for you important to know then you have it.

01:13:44:19 - 01:14:14:09

Marco Schmidt

So when have so much disease then you are Nigeria's words and you are only when you are. And two weeks or three weeks before you die, then this is actually this is actually not genius value for you. The value is then you have it when you're 50 something like this, and how you can prevent an outbreak of the most disease and is yet another prediction model.

01:14:14:09 - 01:14:43:11

Marco Schmidt

So you had to call into action of what you for this other genes but provided us better would enable us to predict in a better. I'd summarize disease but she was in the problem look you know you know that the likelihood is much more higher but there is no treatment. So It's unethical to give you the information that you really have to understand so and so.

01:14:43:11 - 01:15:11:18

Marco Schmidt

So we have done this for from a customer. But you see, this is the prediction I know is that a lot of I call it in this delivered in digital health, the prediction is actually something that has value in our health system because first of all, there is no treatment available. Then it's unethical for a physician should not tell you this.

01:15:12:16 - 01:15:33:06

Marco Schmidt

And the other way around is what courtesy of your prediction, because there's still a chance that this is just nonsense. So you really have to prove it. And to prove it means that you have a prospect of clinical trials in these 20 or 30 years. And then it's very expensive to prove it. Very expensive, and it takes ages to prove it.

01:15:34:04 - 01:15:34:20

Marco Schmidt

I mean, so.

01:15:37:22 - 01:16:09:09

Christian Soschner

A solution or a therapy can always be a product like in the pharmacy industry. So it's usually a pill or it's a liquid that helps patients. For me, I mean, this is one of the driving questions of my life. I would like to understand how diseases develop and a therapy then also can be some lifestyle changes to avoid getting a disease or I mean, the ideal point for me would be leave happy and healthy for seven, eight or nine decades on the planet.

01:16:09:15 - 01:16:37:02

Christian Soschner

And then when life comes to an end, die fast. So to avoid this this long of of a disease so suffering from from a severe disease for two or three decades. And for that's I think it's my opinion it's necessary to understand how does the biological system come from a healthy state up to a to a very long progressing state of of disease instead of just shutting off from one day to the other?

01:16:37:19 - 01:16:43:08

Christian Soschner

And I was hoping that with artificial intelligence, it's it's easier to get a better understanding beyond drug discovery.

01:16:43:09 - 01:17:13:16

Marco Schmidt

Drug Yeah, definitely. Definitely. So for example, type two diabetes and we have extremely, extremely good prediction what it's not. So I think it's 99% accuracy you see in prediction for type two diabetes. So there's a lot of I'm just, you know, that there's a lot of genetic impact and then it's actually your body rates you see in a direct in the diet direct relation to it.

01:17:14:03 - 01:17:48:21

Marco Schmidt

So if you have you have problems with your body weight and you do not do some sports and then you have this impact and it's very likely that you have a which is that you're going to diabetes. It's coming to your question. So this reminds me to a project you also had. So we looked in the genomes of long living people, so 101, another 20 And so they, they check them in the genomes and compared to people died early.

01:17:49:14 - 01:18:19:02

Marco Schmidt

And yes, there was always the idea that this is the longevity gene. But actually what we saw this and longevity genes are, I call it short to the genes. So they are really genes that are for diseases. And if you date early, they are these genes are extremely overrepresented in this in this cohort compared to the long living people.

01:18:20:01 - 01:18:38:20

Marco Schmidt

And this has actually been my takeaway. So I think at some point this education, you have an idea how to live healthy, but if you have a just a normal genetic make up, then this actually you big risk for your life.

01:18:39:23 - 01:19:02:15

Christian Soschner

And it's very interesting to talk about the different facets of artificial intelligence. When asked the first time this this longevity question. So out of personal interest in the nineties, there was almost no answer from science. So think it's just really really way back free for decades and when I look now on the Internet, I think the availability of data is much better.

01:19:02:15 - 01:19:27:12

Christian Soschner

There was some years ago, ten or 15 years ago, I read the study from I think it was Harvard University, where they did a study over more decades where they observed people with a questionnaire. You ask the question, what are your lifestyle habits? And then came to the conclusion that there are four or five factors that influence health tremendously and.

01:19:27:12 - 01:19:59:02

Christian Soschner

Now, before artificial intelligence, we could scale it up. Is there any chance to get the necessary data to get more and more insights into these mechanisms or I mean, in Europe we have our GDP rules, which protects the data. What are the limitations of regulations that limits the development of artificial intelligence to interact discovery on one hand, but also on the other hand in getting more insights and understanding of the development of diseases?

01:19:59:02 - 01:20:19:20

Marco Schmidt

Yeah, are two things. So talking about the GDPR, I think it's just an excuse for bad research. So I it's, it's very hard to answer now because it's actually research. So when you look into the legal text, you see that when it's researched in GDP, it's it's not so important. So you can use, you can use the data.

01:20:20:12 - 01:20:47:03

Marco Schmidt

But I always see this people conduct clinical trials or genome wide association studies in Germany. And when you look a little more closely, then you see that the book is not probably done, but it can be reviewed easily when you see the data but this and say, no, there is this data protection issue. That's why they do not disclose the data.

01:20:47:10 - 01:21:07:24

Marco Schmidt

So they keep a little bit the fact that they have not done it properly. So this is a little bit my opinion on this because all people talk about this when it is research, then it is a large group when you look at medians and you can text. So this is actually not this is actually not a problem.

01:21:08:13 - 01:21:34:22

Marco Schmidt

And to your other point, with the longevity, longevity, there's a lot of movement now in the field and there is this famous phrase and I can also on our side here. So, I mean, you look at I use Silicon Valley billionaires now. So when they were young, they wanted to be rich and they're rich. They wanted to be on.

01:21:34:22 - 01:21:57:24

Marco Schmidt

And that's why you see now a lot of money from people. For example, Elon Musk, They're heavily invested in the team. It's it's really incredible. And you see a lot of things now upcoming, my opinion, because for them, money is not an issue. They have a cool life and they want to stay alive. They want to enjoy their life.

01:21:58:08 - 01:22:35:16

Marco Schmidt

So they are they are they have actually unlimited resources from from the money perspective. Yes. I'm pretty sure that we see lots and a lot of data to the and I'm pretty sure they've also put this down publicly because I haven't mentioned this before, but this is really important. So if your data is in your silo, then the use of the data is limited by you, by yourself, but then you put the data into the public, then you really see that other people come these new ideas, this new methods, this new technologies, and they really make it useful.

01:22:36:01 - 01:22:58:13

Marco Schmidt

And it's true. And they also find in this taking your data, when people always say, Oh, that's only mine, and the GDP or whatever, then you can make sure that there's something wrong with the data because then the data is proper, it's created and so on. You invested so much and you are your person. You want to see, you want to see the results out of the data.

01:22:58:13 - 01:23:20:05

Marco Schmidt

Then you're more open to say, okay, let's make it public to all the people outside of the good, because then we can I just can check on balances from other sites that they are good results and we can use it in the end. I want to this longer, more healthy. This is my actually goal. And as I said from the start, from the Silicon Valley billionaires.

01:23:20:17 - 01:23:28:01

Marco Schmidt

So they do not argue as money that when you argue about the boards, allow them to.

01:23:28:01 - 01:24:02:06

Christian Soschner

Now this was also more of selfish interest in the nineties. My message was I want to stay healthy, I don't want to get sick, so what can I do to increase this chances of staying healthy longer? You never can avoid any accidents or I mean, sometimes life happens. But I think also that said, as a huge impact. And I'm curious to to hear also from people like Elon Musk and Bill Gates and I think also Tim Cook, it might be something to bring into the area what conclusions they draw when they put really big, big numbers behind it.

01:24:02:07 - 01:24:27:09

Christian Soschner

If increased in anger, May is also someone who is investing heavily in that space. When we look at the situation right now that you mentioned with the availability of data, wouldn't it be helpful to access also on what you call better data of data from from studies that failed to increase the know how?

01:24:27:09 - 01:24:31:11

Marco Schmidt

You know, that's I mean, that data is I call it badly created data.

01:24:31:23 - 01:24:34:02

Christian Soschner

Okay.

01:24:34:02 - 01:25:11:00

Marco Schmidt

So for example, when we started companies, we had only access to high quality data sets and then we had to build our own quality control unit. And sometimes it turned out that 20% of the data was useless. So because we extended quality control, this data could not posit. Good example was and so you have the genomic information and then you have the phenotype in this type of so there's supplement, you look in the genomic information, this is actually a map.

01:25:11:15 - 01:25:49:14

Marco Schmidt

So then there was in the discussion. Now probably it's intersectionality, but 20% come on, nobody continues to turn to this happens 20, 20% of the time. So this a huge and this is just there's only a sign that it was not properly done. So new. But anyway so would you have done a biotech system And we took these badly created data and we created the model and our own ways of testing all our standard quality control.

01:25:50:04 - 01:26:04:11

Marco Schmidt

And from this data sets or roughly 80%, But then you is this is useless. Is this useless science useless institute for 20 20% this and 80% is used for you can then go on.

01:26:04:11 - 01:26:33:17

Christian Soschner

So to improve drug development, if I sum up the conversation until now, in early stage development, there are two problems to solve. One is, getting access to data, and the second one is the creation of the data and companies like you asked. Then when they really sit down and do the second step and to reach the data, you have 80% of half the data you get that is useful for further verification of solutions.

01:26:34:07 - 01:27:05:22

Marco Schmidt

Are only in the beginning, in the beginning. But what you see now is that when you look at the Biobank data, which is becoming more and more available, was the time this is now really excellent. So you can have proper quality control units and so on, so in the beginning it was mostly useless, but they have really learned sort of, for example, the back of trust test and invested massively in the beginning in this area.

01:27:05:22 - 01:27:36:02

Marco Schmidt

So so that Case Control Consortium, for example, and the state in the results are published in nature, so that a lot of this a lot of better quality in it but to become trust. So they paid it, but they also okay look what are the findings now five, ten years later? What would we have done wrong and what we should what we can how we can avoid in the future?

01:27:36:12 - 01:28:09:24

Marco Schmidt

And these learnings were taken especially by UK Biobank and Genomics in this way in the UK, and then it's used in the normal biobank data. So I think be much a really could be. So I know people always complain about the quality data, but in my opinion nowadays it's a little bit also kind of we call it baked setting discussion because people do not understand cyber reasons and how you can provide value and so on.

01:28:10:11 - 01:28:12:11

Marco Schmidt

So always say always a data city. You know.

01:28:14:04 - 01:28:37:02

Christian Soschner

The I mean, for me as an economist, quality is a function of capital. So I mean, you mentioned before that studies are cheaper in Europe, but when we have less capital, we can do as much as as steeper pockets can achieve. And also Wellcome Trust, I mean, a completely agree with us, the Wellcome Trust, they're doing a lot of good things and bring research and move research and development forward.

01:28:37:20 - 01:28:53:07

Christian Soschner

You mentioned Biobank. I'm just curious, maybe it's a stupid question, but I ask it anyways, in your opinion, to get more data that moves aid track development forward, how many biobanks would be need in the world to increase quality and accessibility of data?

01:28:53:20 - 01:28:55:20

Marco Schmidt

And we need only one biobank.

01:28:57:12 - 01:29:01:06

Christian Soschner

But how? How much capital and how much resources to get on this?

01:29:01:07 - 01:29:28:14

Marco Schmidt

In my opinion, when you look, for example, at the standard in the development countries, for example Western Europe, in my opinion we are talking about let's say a peanuts because a lot of what you know, do it is done by the daily routine of the hospitals. The only way is known to documented and put on top for example genotype testing.

01:29:28:14 - 01:30:05:16

Marco Schmidt

So genotype testing it costs €25. €25 is genotype testing. It's when you run use of genotype chips from Illumina, for example. So it costs you also with the quality control and plus in Bioinformatic processing took you or it takes your 25 years. So it's incredibly cheap and there were human genome sequencing is now below 1000 and last time it's about a red 100 U.S. dollar.

01:30:05:16 - 01:30:43:13

Marco Schmidt

So this is so including, you know, and the people who cover the data. This is the physicians and nurses in the hospital. And I think it's only important to keep an eye on sick people, because enough people already available that we have information system that control and then you can manage this. So in my opinion, it's they are just only excuse this was GDPR and that there's not enough money available.

01:30:43:20 - 01:31:03:11

Marco Schmidt

So what you can save which can really bring to the people and to the patients, I think it's incredible. But the investment, my opinion and just the idea you could be can quality a specific concept of entrepreneur estate. It's it's almost nothing because they pay already for this.

01:31:04:07 - 01:31:13:24

Christian Soschner

So we have the data we just need to structure the processes properly so that the data is created and accessible for scientists to fund research.

01:31:15:02 - 01:31:55:17

Marco Schmidt

You know, and this and, and my opinion and this my learnings from the last five years and then you bring the data to the public or make public, then you also see that a lot of physicians probably don't do their jobs properly. So you have said also kind of a quality control all people because you see everything and then you see that the person is diagnosed, for example, for a specific disease, but then you see all these biochemical parameters in the data as this kid cannot be this disease.

01:31:56:05 - 01:32:25:02

Marco Schmidt

And I think that's a lot of this. Yes, this a real problem and said to this and they do not want to be monitored in this way. And I think that this is actually more the problem than GDP are so expensive to set up a big biobank, especially in Europe, because you are not ready for this and you pay enough for this can do it easily because it's so expensive.

01:32:25:23 - 01:33:12:19

Christian Soschner

And I completely agree. You think the health care system in the United States and in Europe is far too expensive anyway, So it's it's fine and say we're going to wait interstate in the states. We have it currently. And when it just sum up and imagine, try to look more at it from a science fiction of view. Again, everything that we can do that improves diagnostics, for example, to diagnose disease is idea or speed up track development processes that bring products to the market much quicker to cure diseases, and also on the longevity point of view, I think this research is also very valuable because it helps people to understand how they can stay healthy

01:33:12:19 - 01:33:24:01

Christian Soschner

much longer, drives down the expenses for the health care system. And this is one of the big problems that Europe and the United States currently have. And artificial intelligence plays a huge role in that now.

01:33:24:03 - 01:34:05:13

Marco Schmidt

But you see a huge difference in the population, in my opinion, because you have some well-educated people. They care about their health and they are they are they're they are willing to invest in the the antibodies. And then, on the other hand, used to see a lot of not so well-educated people. They just want to say, look here, I still want to drink my beer in the pub and I still want to enjoy my last and this unhealthy this is unhealthy lifestyle, but just deliver me the plural I take to put it every day.

01:34:05:13 - 01:34:33:01

Marco Schmidt

And it's it's you know, but I think you should talk a little bit more about the genome. So I sort of pointed out here because I see and know and I think that this is a region way we talk. So as human genome. So it was sequenced more than 20 years ago or so in roughly 2000, 2000 200,000 published.

01:34:34:03 - 01:34:57:21

Marco Schmidt

But we still now see more and more impact of the human genome so as to as a comma in the street. So as I said, your Ohm's Law, for example, additional breaking your Ohm's Law from 2010 and so on and I see here the real value of actually just like the artificial intelligence, because there's no definition of intelligence so how you want to define artificial intelligence.

01:34:58:12 - 01:35:45:24

Marco Schmidt

So I would like to call it predictive statistics and also causes statistics causes inference. So and this is actually the opportunity. See, because as I said before, the big problem is just the efficacy of your truck. And then you look from from a mass perspective to it. So efficacy and causalities actually is the same. So in 2019 there was a Nobel Prize in economics Science Institute, for example, and she had this brilliant idea of conducting randomized trials to get an idea of bizarre development.

01:35:45:24 - 01:36:19:13

Marco Schmidt

It really works in India or not. So and she conducted a prospective trial so selected random in a randomized allocation, a group of poor people in India who receives development aid and the others do not. And then she could compared and this is extremely this is actually the gold standard that they from medicine to economic science but it is so expensive is so expensive to conduct a clinical trial.

01:36:20:04 - 01:36:46:17

Marco Schmidt

This is also we had to India in the beginning of our conversation, our clinical trials, the system where you have to find the patients and you have all this not enough patients recruitment is slow. Then you have, you know, so you assumption it's the reality. I call it the sum of your intuition. Contrary up, they do not match what's available patients and so on.

01:36:46:17 - 01:37:09:19

Marco Schmidt

And in economic science. And it was very interesting that twist and so last year there was a Nobel prize said for cause entrants and so yes a problem the economic sense that you cannot run trials prospective trial it's all the time so they looked in ways how can use observational data and prove causality out of proliferation of data.

01:37:10:05 - 01:37:40:02

Marco Schmidt

And then they open up the completely new of the division of statistics and this is causing insurance. And to point it I see you know, is the biggest issue for artificial intelligence and machine learning because insurance because building in bring causes insurance. So we just want to check for causality in the data and is often why as this I call it, the three point being so you have your your instruments, it's actually your genetics.

01:37:40:14 - 01:38:04:08

Marco Schmidt

And then then you look for a trade in for the disease. So Very famous, this cholesterol, for example. So you have genes that can increase cholesterol in the human body. But on the same way, the increase the risk of heart attacks. And then, you know, can what if you have genes that just only increase a little bit of cholesterol, they only increase a little bit.

01:38:04:08 - 01:38:38:03

Marco Schmidt

Heart attacks increase a lot of cholesterol, also often a lot of heart attack. And this is the relationship you can the model and then you can say, okay, this is this is cause of such a causal causal relationship. And to apply this now you combine this with I'm predictive modeling, so you have prediction models where you say, okay, I take no the point where I know this is actually I think it's causing I took it out of my model and then I look for my prediction of my mother.

01:38:38:13 - 01:39:10:03

Marco Schmidt

So here I see a lot of potential and the genome gives you the link to the biochemical level because you actually see whether you have this either in the genes or not. And if you have seen this biobank 2 million, for example, genomes fully created and this phenotypes, then you can mimic actually whether some alteration or some modulation of a specific target gene can work or not.

01:39:10:13 - 01:39:13:23

Marco Schmidt

And he see a lot of potential.

01:39:13:23 - 01:39:53:23

Christian Soschner

So can we can mimic to test that does mimic mean that this study is simulated. Yeah so so not because I mean I just remember the conversation you had with Emily Mitchell about decentralized clinical trials. You mentioned a problem that of getting enough patient patients, a large number of patients into a study is a science by itself. So it's not easy to do a 45,000 patient study in a phase three setting like, but they are not even getting 200 patients sometimes in reality is is is quite stretch and difficult.

01:39:53:23 - 01:40:15:13

Christian Soschner

And I saw some cases where it can take 5 to 10 years to get this basic patient population that really is willing to participate in a and one solution is decentralized clinical trials. And it I really understand your ads that we are talking about. We had deficient intelligence about similar think entire trials in a computer.

01:40:15:24 - 01:40:39:08

Marco Schmidt

And so what I so what I would like to point out is that you have these two things when you want to develop a drug when this you need elegant and you need to safely and and in opinion there are a lot of methods now available where you can test for hand and see whether it's safe or not.

01:40:39:21 - 01:41:16:11

Marco Schmidt

And the other big problem remains the efficacy. So you will see an effect due to the modulation of the target you choose. But what you can do now is you look at the genome course, some of what you do is too many people and then you look for alterations for the specific gene. And so we have done this in the past so we were interested by people that have a severe root cost and the other features only have mild, moderate and the applied in this cause inference methodology on these two chords and what you should got out.

01:41:16:11 - 01:41:46:00

Marco Schmidt

COVID 19 patients had higher neutrophils, much more high immature features than people with mild moderate cases, but moderate COVID 19 cases. And we could really saw or we could really demonstrate this was a prospective regression model. In job one analyzes propensity score this tool with a lot of mathematical methods. We were able to to show that this there's a strong evidence for the neutrophils.

01:41:46:00 - 01:42:10:21

Marco Schmidt

And then we went on to look into the genetics of some natural foods, and then we saw that genes that have been significant for neutrophils, for the number or for the Newtons or for a neutral for cyclone in general. And so you can then use the entire population are from the cells and people. And then you also use smaller signals because the population is much, much more strong.

01:42:10:22 - 01:42:43:02

Marco Schmidt

It's much, much bigger. So you see the yard effect. Size and sets are smaller than usual. That's why when you look just only at COVID 19 patients. So in the literature, you will be 1000 or 5000 cases available and these populations are too small to catch up all these very interesting genetic variants. And this was this kind of trick that these you can see artificial intelligence methodologies.

01:42:43:11 - 01:43:29:13

Marco Schmidt

They enabled us to find a causal truth as well as the nutritious. And we can look in the entire population and look up what kind of genes modulate which for so-called and we ended up with CDK six. So CDK six is an enzyme which is responsible for the cell cycles originally. But as an extremely strong impact in neutrophils and especially in specific, I'm in a specific job that's unusual for two as part of immune response and this is so quiet and it wasn't so this is a special form of apoptosis and these little threads just blew off and they release so quite neutrophil extracellular traps.

01:43:29:13 - 01:44:05:15

Marco Schmidt

And what you see then I would just add all these drugs that have been tested successfully so far today. So for example, this is a financing in 2020 on certain components, and they tested successfully against severe COVID 19 patients and then another cancer drug so far was recently also tested persistently. And what we then saw and in vitro experiments, this was a partner from the University of Bristol.

01:44:05:22 - 01:44:45:14

Marco Schmidt

They said all these components they inhibit actually is this net neutral for the ActionScript formation. So they chose us. So what I want to tell you is that it was just a simulation, but then you do it. Now just this you are this the population data you have, you are so close to clinical trial. So that's why it makes sense to simulate it first and using the naturally occurring alteration in your genome to see whether they are impact on specific traits which then are causal for a specific disease.

01:44:46:01 - 01:45:04:14

Christian Soschner

So your so your model was able to rightfully predict what's going on in the human body. You've a big problem is and all the therapies that came later to the market then basically tackled exactly this problem that you predicted.

01:45:04:15 - 01:45:29:23

Marco Schmidt

Yeah you can say you can say the drugs and this is nobody, but it's now in revision application. You could see in our analyzes that of all these kind of drugs which have an impact on the severe COVID 19 status, to have an impact on the specific stone tools on the neutralizes so they inhibit or the reaction of the human body.

01:45:30:06 - 01:45:41:10

Marco Schmidt

And the other reaction is mostly it's you can really narrow down to the neutral stance, but this is what you can see, which you could see and in the data.

01:45:41:22 - 01:46:09:23

Christian Soschner

So when I would try to thinking of commercial applications to one application, be that you have a tool to tend for the pharma companies that before to go into clinical development that the check they are their compound, their asset whether it detects the problem that you predict and then is included in the decision making process, it makes sense to move with that compound for or it's better to switch to another compound.

01:46:11:12 - 01:46:19:08

Christian Soschner

So what would it be a possible commercial application of this?

01:46:19:08 - 01:46:46:03

Marco Schmidt

Is this actually what we do is so we offer this in the service model mostly we call this emerging biopharmaceutical companies. So as a little bit the reason that big pharma companies, they have small departments and they think they can do this better. But anyway, no comment from last night. So an excuse is the reality is actually this whole cycle with the pharma big pharma companies, it's it's take you a one year or longer.

01:46:46:22 - 01:47:19:17

Marco Schmidt

So we working mostly with smaller companies and then this second it's much more it's much more faster which is for us more important. And what we do is we call it's interest into the clinical trial that we provide not only an idea, but it is efficacy. So is this what we have done with this qubit, for example? But what you can also check is then you have a molecule already in the market and you interested in a new indication for example, especially in India, rare diseases because they don't sell.

01:47:19:20 - 01:47:48:06

Marco Schmidt

And as I said, even this a more patient group we can determine the cause of traits and then you can go back to the bigger population and then and I call it simulates the cost on these disease patterns, the patients and the big population and look whether they should reflect on specific genes or not. There you are original molecule we know that is this is it can what are they traits or not.

01:47:49:00 - 01:48:13:20

Marco Schmidt

And the last one I think it was also interesting I said before safety can be checked in the early face. That's right for some in animal models for example but due to the fact that you only test the components only for a specific period of time, in the beginning, it's just only you want to move on. So you do the test for one or two months, for example.

01:48:14:12 - 01:49:00:11

Marco Schmidt

But sometimes when you have, then you have the treatment which has to be given the way just to be administered for use, for example, for chronic diseases, especially then you really have to check whether the inhibition or the ambition or the mutilation due to your molecule to a specific does not induce. And other agencies are example. In the past we saw that it was inhibition of a specific channel could be to see for example but this was what was not was really a danger.

01:49:01:02 - 01:49:29:03

Marco Schmidt

But you have to lower than before because you can think about how long the treatment can be so that you say, okay, is this is just only for the indication. So this can only be done. It can only be given for a specific period of time and then you have to check the patient with the symptoms regarding in this area or not or are there, for example, genetic variations, which makes the even stronger.

01:49:29:03 - 01:49:46:16

Marco Schmidt

And whatever more changes, the higher likelihood that this adverse effect would occur. So there's this idea, this few things you can do actually know the status and you should check for the efficacy. You can check for new indication and you can check for target toxicity.

01:49:46:16 - 01:50:25:05

Christian Soschner

That's that's awesome. So drug development simulation I think isn't borne probably most of it in clinical trials to find direct patients. I mean, you mentioned Corbett and I was thinking that we were speaking to a senior about this, this problem. So I mean, from the information I have, which might not be right, but my perception of the Corbett vaccines is that they work really well for severe cases and don't have much effect when people are not at risk to get the severe disease compared to compared to the situation of the chest.

01:50:25:05 - 01:50:56:12

Christian Soschner

See to Totalement get over it in a couple of days. So from from from the patient perspective, but also interactive development. I think the interesting thing for trials with vaccines tend would be to find the patients that the highest at risk for getting the severe disease. So also see the effect in a clinical trial setting so that it works or not, the vaccines might not be the best example because for the dynamics of vaccines to to to protect the health population, but benefit cancer, for example, that's your solution.

01:50:56:12 - 01:51:26:03

Christian Soschner

Also help to find the right patients for our study. I think just when when we get a lot of capital in a clinical trial and we take the wrong patient population and theoretically the drug would work very in a different kind of population than we actually caught it in a trial. When such a clinical trial fails because of selecting the wrong patient population, I think it's really hard to bring it back.

01:51:26:07 - 01:51:28:22

Christian Soschner

So can your solution help us? In many cases.

01:51:29:10 - 01:52:11:14

Marco Schmidt

Yeah. So we had another project team, the company it's Lactoferrin for Origi's. This disease is quite a disease, so it's an auto inflammation disease. When we have sustained for problems with your eyes and the overall idea and then the introduce to the component plus the disease patients, I would imagine so lactoferrin when you're use it so lactoferrin this able to bind.

01:52:11:14 - 01:52:49:23

Marco Schmidt

So they said that some this would be then an ironic thing and bizarre to disease patients. So is it what was originally claimed? And then you looked a little more closely it's to be run our cause modeling on this and then you forgot oh it's it's actually based on the when it was here it's so it's a specific Yeah you can say immune cells that are going into the eyes and then they start infiltration and the infiltration is then the cause of the vasculitis, which is then the real of the resistance is.

01:52:50:10 - 01:53:25:14

Marco Schmidt

So what you could figure out is that you always interfere. Excuse the point. It's this is a problem and then you can use it as an endpoint through clinical trial instead of the IRA. So this to IRA, because I and has actually no clue and our cause and what would be CMAG population and so on. So when you go into a clinical trial you just look for you as interference instead of the IRA, and then you have a better endpoint because you're already proving the task causality between the initiates and the outcome of the disease.

01:53:25:20 - 01:54:07:18

Marco Schmidt

And notice, Ira, when you have just assumptions to this, your your your model was the IRA in Ireland you see everywhere you know some so to say in this day and then it's very likely that you you because you have just look for the wrong for the wrong parameters in your data and the three astonishing courses, then you can choose better endpoints and you can also this is a bio statistical problem that you then also can select or you can shoot for your if you are number of patients for a clinical trial, because what you want to see much more clear for you.

01:54:08:08 - 01:54:38:01

Marco Schmidt

And then you only need you need so much patients anymore and beginning with just one a group. But if you show this effect in the smaller one, the place you will get should be significant in this. This is my opinion because Thailand patients this this is this is the next big claim after you predict efficacy in genome wide association data and population data.

01:54:38:01 - 01:55:04:05

Marco Schmidt

So you have you still have to prove it prospectively upon the regulators. But if you for example, for a rare disease, you have to recruit 1000 patients. So just like, okay, forget it. If you remember, find 1000 patient, you have sites all over the world and you you are willing to pay incentives for patients to join. It's it's so it's really tough.

01:55:04:12 - 01:55:21:08

Marco Schmidt

And you'll see also another point we really have to see, even when it's about just cause the and on synthetic control arms and synthetic control arms they are fantastic thing. I don't know whether you've heard about this.

01:55:21:09 - 01:55:22:17

Christian Soschner

No, not yet. What is it?

01:55:23:16 - 01:55:55:14

Marco Schmidt

So the richness in oncology, in oncology is so that for some tumor diseases there are some drugs available, I call it in the space, very, very simple. So then, you know, have a new and you want to come this the new drug Mr. you see this a much better drug to the market you have to conduct a clinical trial and it did so that in the standard setting we have six placebo controlled randomized clinical trials.

01:55:55:14 - 01:56:20:21

Marco Schmidt

So you have your population and some we receive just a placebo and the others we receive your drug monitoring. But to receive a placebo for a tumor patient means actually is test. So and then you have one unethical clinical trial because half of your population will die for sure because they are not treated and that is the poor treatment available I call it.

01:56:20:22 - 01:56:47:04

Marco Schmidt

The poor are not so good treatment available, but if you don't give it to them, then you should really be the immediate point, you know, then the idea okay, I do is a clinical try against the standard of care to support treatment. And then they had sometimes the problem that this poor treatment they are off label so they are not really officially given and so on there was not so become more and more problematic.

01:56:47:10 - 01:57:16:14

Marco Schmidt

And then you had the possibility that you took retrospective data of patients and you just model. Now the progression of the tumor grows. It just can be really done in a simple Excel sheet. So it's just lots of stuff so much sophisticated. And to be honest, it's very simple. But you need the data of the tumor patients and you simulate the way how a tumor is going without the treatment.

01:57:16:21 - 01:57:55:23

Marco Schmidt

And you just only have one arm of mutations and you see the patients receive your medication and you test against the synthetic clinical control arm. It's just established, I think last 5 to 10 years, most improved quality and these synthetic clinical to control arms. So this is also called the insurance methodology of develop. Actually also hear from the one guy who received the Nobel Prize last year because he wanted to he wanted to process immigration from Cuba to Miami does not have an impact on the unemployment rate in Miami.

01:57:56:16 - 01:58:21:08

Marco Schmidt

But then he wanted to compare Miami with another US city. And then you realized this actually, you know, and so then and then you created this synthetic control on that. You said, okay, so this is my Miami without immigrants. And so and this now also now coming back to clinical trials. So this idea of causality and I see also now a huge improvement.

01:58:21:11 - 01:58:45:10

Marco Schmidt

So a lot of things now going on that you can decrease in number of patients in the clinical trials. So and then people know that this new placebo controlled randomized clinical trials. So it's not a chance to just to receive the placebo because this is actually also some kind of jeopardy for these guys. Then they are also more willing to say, okay, this is new treatment.

01:58:45:10 - 01:59:02:08

Marco Schmidt

And I, I read your papers and it looks good for me. And the simulated that can be hard to see after the clinical trial. Then they go for it. Otherwise, people say, well, given that is the 5050 chance of what we saw in the past.

01:59:02:22 - 01:59:23:16

Christian Soschner

But what about it? Biosimilars wasn't the reason to invent that. We planted clinical trials with placebo control to to reduce bias. Is is that a problem then? I mean when people know already that there will not be a placebo, does that jeopardize the results of the study in your opinion?

01:59:24:03 - 01:59:56:00

Marco Schmidt

Yeah. Yes. But do you have to as a synthetic clinical trial, you have to understand properly. So he should send, he will interpret. But I would like to I would like to tell you one little thing about the thing you have seen in the past in clinical trials because of the jeopardy for the patients. So in the past, it was seen more and more difficult to run these clinical trials in Germany because people say, okay, there's a 5050 chance I take actually nothing.

01:59:56:08 - 02:00:31:03

Marco Schmidt

So nobody will participate. And we saw that a lot of movement of clinical trials and told by Ukraine and Georgia, for example, because they so poor systems that patients say, okay, so the standard of care is nothing. So I participate in the clinical trial. Then after 50% chance that it gets the treatment. So this is what we have seen in the past, and that's why I see this distinction, is it's not a statistical message.

02:00:31:14 - 02:00:39:05

Marco Schmidt

So really, I call it a triple effect. And you know what? In economics, you also for patients in lower income countries.

02:00:40:02 - 02:01:00:09

Christian Soschner

Before we come to the last two questions, I want to check back you from the study. Right. You said possible for stimulation of synthetic controlled arms. Excellent Enough. At what point is enough? Is is is is really excellent. I mean, Microsoft, the one the tool that we have attempt.

02:01:01:15 - 02:01:39:06

Marco Schmidt

So in the beginning is was really excellent. I know that the big provide us awesome to the clinic at once and you will not always hear me say they have is they are also you was sophisticated sophisticated algorithm but in the end you can also do machine learning was excellent. I know I know a lot of than you know I guess probably it will end up not as the system but it's possible you can do it and believe me I'm big organizations on Microsoft excellence to the standards.

02:01:40:02 - 02:02:14:09

Christian Soschner

Yeah I know. It also other areas but it's my experience is 20 years ago it wasn't extracurricular activity at understanding. It was in information technology. It was not in the industry, but very often when there was a salesperson saying it's an automated process, the data is handed over. Ultimately it's the reality was that someone freelancer was sitting somewhere in the world taking the data from one database, transform it, transforming it manually in and uploaded to the next database.

02:02:14:09 - 02:02:21:01

Christian Soschner

So this is was very often the reality and it's good to hear it actually still useful today. So Bill Gates did a good job at the end of the day.

02:02:21:18 - 02:02:27:11

Marco Schmidt

We are the principal mass. You know what? You can change. You know.

02:02:27:11 - 02:02:29:08

Christian Soschner

You can live in the change of.

02:02:29:20 - 02:02:33:24

Marco Schmidt

Government. It is becoming bigger and bigger. It's it's a problem for sure.

02:02:34:07 - 02:02:45:19

Christian Soschner

But for it's not. But for a starting point, it's still one of the best tools to use. I mean, that's that's a bit let's look at it from the positive side. It's one of the best tools because you can easily model something.

02:02:46:14 - 02:03:17:10

Marco Schmidt

But you have and you look at specific examples. So you look at the tumor growth. So you have trauma patients from the past. You monitor, what, about five, ten years, sometimes sometimes shorter, because survival rate is not so high. So you try to make these patients to one average patient interest. First of all, you then you want to condense all these people to one person.

02:03:17:10 - 02:03:21:16

Marco Schmidt

You need as a towards an excellent for sure, but they should come to the stage two meeting.

02:03:23:09 - 02:03:29:07

Christian Soschner

But if you mean the critical message for me in there is it's not difficult to get started on this journey as an organization.

02:03:29:07 - 02:03:31:05

Marco Schmidt

So it definitely.

02:03:31:05 - 02:03:51:12

Christian Soschner

Doesn't it doesn't need tremendous amount of investments upfront. So the simplest thing to start is you get the data scientists use X to get started right now. And then when you progress your organization on the journey, you can always invest in more so for sophisticated solutions. But the first step is pretty simple.

02:03:52:03 - 02:04:16:01

Marco Schmidt

Yeah. So this is actually, I would like to call it We are living in fantastic times because everybody can now do fantastic research together. This this is computer from home. So because if you need more computation power, then you can use cloud solutions. For example, in a lot of companies, I see a virtual now in the development space.

02:04:16:02 - 02:04:37:02

Marco Schmidt

So do all this kind of validation computationally and then they select drug molecules, buy them, test them and contract research, and then they bring it into the clinic. So this is a lot what you can do. Computationally so, yeah, as it said. So software is getting good.

02:04:37:02 - 02:04:56:17

Christian Soschner

I mean, these are fantastic words at the end of the podcast that that you said we are living in the best times ever. When I think back to the eighties when I had the question, it meant traveling the whole world to get an answer from an expert. Today we have a Zoom call. I can ask you what my question is about artificial intelligence in track discovery.

02:04:57:07 - 02:05:17:20

Christian Soschner

I can read your posts on LinkedIn where you give away a lot of information for free. And my last two questions to you on this podcasts are is the one topic that you would like to talk about at the end of the podcast that I didn't ask you questions about it.

02:05:17:20 - 02:05:43:19

Marco Schmidt

Yeah, for me it was just only important to highlight that in the story. We can wrap it up here, so wrap it up. The important part is we have all the hype this, all this ligand ligand ability, this Alpha four that, you know, the protein structure and can do molecular modeling in a better way. But actually the revolution is done now in drug ability.

02:05:43:19 - 02:06:11:01

Marco Schmidt

So how we can select the right drug targets? How we can find new indication, especially for often diseases, is would be tremendous to imagine this. So now you're checking mutual menu in your simulators and then you get a good model for just 510 patients. But why? And it's much easier to convince regulators to conduct a clinical trial with three or four patients.

02:06:11:01 - 02:06:26:10

Marco Schmidt

Then you have this evidence found in big population data. And of course, this has an impact on how we conduct clinical trials. So how we think about causality change in our clinical trials, tremendously.

02:06:28:02 - 02:06:54:23

Christian Soschner

Great points, great points of joy. And my final question to you then is when someone is after listening to the podcast review to get started on the journey and would like to include your team in they Are studies, or on the other hand, an investor would like to explore ways to invest in your other ideas that you are about spinning out into into separate companies or do it in your company.

02:06:55:05 - 02:06:57:23

Christian Soschner

What's the best way to reach out to you?

02:06:58:16 - 02:07:26:10

Marco Schmidt

Yeah, so just drop in the immune saw. I'm always keen to to talk to people, so it's always interesting to learn more because in the end it's the people that business. So if we have a problem in the area that you're not you know you're not sure about the efficacy of your company, you look for any indication or you want to check the toxicity, just drop me an email and people have a look.

02:07:27:01 - 02:08:02:19

Marco Schmidt

And if you are general interested in this topic of how in our drug development change from more reductionistic to realistic find, I can also recommend my book, Chemical Biology, where I describe two parts how now the word industry is changing because of this new technology is in. Because when you look at crispr-cas, for example, by artificial intelligence. So that was not the 15 even ten years ago.

02:08:03:04 - 02:08:07:11

Marco Schmidt

And it's completely changing those industry. And this is extremely exciting. No.

02:08:08:05 - 02:08:17:08

Christian Soschner

That's true. That's true. With your consent, there will post your email address, your LinkedIn profile, the link to your book on Amazon. I think it's on Amazon also. So it's easy to.

02:08:17:08 - 02:08:26:16

Marco Schmidt

Yeah, you can, you can book it everywhere. So it's so the publishing all this is the richest but also yeah so.

02:08:26:24 - 02:08:47:16

Christian Soschner

I will post this links in a description to the podcast something people can find you. Michael thank you very much for this insightful conversation about artificial intelligence and drug discovery. I learned a lot of new things. It was amazing talking to you and listening to your answers. I wish you and your team happy. We can use good luck for your future and development of your ideas.

02:08:48:14 - 02:09:09:24

Marco Schmidt

So much and I thank you for having me here. So it's so I always really like having people to ask questions, you know, because you need to be on the right track, you know, And as I said. So for us, it's really important to bring these new technologies really into the field because we see a lot of value for people outside.

02:09:09:24 - 02:09:21:00

Marco Schmidt

And it's so important to let people know about this change as we have here. And that's why I'm so happy. I could tell a little bit about us and our technology is here today.

02:09:21:14 - 02:09:34:14

Christian Soschner

Thank you very much and have a great weekend. See you soon. Thank you. Bye bye. So, so did you like this episode then? Please, please leave a positive comments and like have a great stay.