Faster drug development through genomic data
Laura Reich Diez
The development of new drugs is complex, expensive and risky. Around 90 % of clinical trials fail because of the right selection of participating patients or because the drug target turns out not to be central to the disease mechanism, after sums in the billions have already been invested.
That's the problem a German startup is tackling. Berlin-based biotx.ai has developed a self-learning software platform that can identify multi-causal relationships and complex patterns in biomedical data and use them to develop new diagnostic tests and therapies.
In an interview with 5-HT, Joern Klinger, founder and CEO of the startup, gives us insights into the biotx.ai technology and the advantages it has over traditional drug development. "It is problematic to predict which drug will be successful, and which will not. So, the question arises: how can we guarantee that an active ingredient will also be successful?", explains Joern Klinger. "This question can be answered at least in part with genetic data."
Better prediction of disease progression based on genomic data
Sequencing a person's genome reveals their genetic signature. This, like fingerprints, is different for each person. "It can reveal patterns such as body size or eye color, but it can also reveal which genetic diseases a person is susceptible to," explains Joern Klinger. "The beauty here is that you can detect diseases very early if you take this genetic approach. That has the advantage that you can also treat them early.
Take genetically caused high cholesterol, for example: the risk of suffering a heart attack in this case is about 3-8 times higher than in the case of elevated cholesterol levels, which are attributable to lifestyle. In the case of genetically caused high cholesterol, certain deposits can form in the arteries already in childhood. These deposits remain throughout life, which means that by the time people with genetically caused high cholesterol are diagnosed with elevated cholesterol levels and treated, it is usually already too late, and the risk of heart attack is significantly increased. However, if the disease is detected in childhood, when the person is still in full health, it is possible to intervene early. This is one of the great advantages of genomic data," Joern Klinger summarizes.
The patient benefits through better prediction of disease progression and early treatment. Another example would be Alzheimer's disease. If the patient learns at an early stage that he will develop Alzheimer's, he can take precautions. In addition, pharma and healthcare benefit from the fact that patients can be better selected for clinical trials, and these are significantly streamlined.
Genomic data and the wide data problem
"Genomic data are not big, they are wide. With wide data, where the sample size (e.g., the number of patients in a study) is much smaller than the number of variables (e.g., the number of variants in a genome), the statistical problem of multiple testing arises. Standard statistical models have difficulty distinguishing signal from random noise, especially when testing complex patterns rather than single variants. So, in summary, genomic data have a problematic structure. Thus, they are not Big Data in the traditional sense, but can rather be classified as Wide Data. The whole thing is best illustrated by the following example: Big Data could be Twitter data, for example. There are billions of tweets, but each tweet has only a small number of characters and thus little information.
Accordingly, there are many examples and one learns something relatively simple from these examples. Wide Data is exactly the opposite of this: there are few examples available, but each individual example contains an enormous amount of information. Clinical trials in drug development sometimes include a few thousand patients. There is an enormous amount of information available for each individual patient; the human genome of each patient already contains more information than there are people on the planet Earth. And that's exactly this wide data problem: the algorithms you knew before work well with Big Data, but now you have to rethink."
Traditional drug development and its disadvantages
As a rule, the development of new drugs is a laborious process with a very uncertain outcome. "Traditional drug development has three major disadvantages, which anyone working in this field will also always list," explains Joern Klinger. "One is the time aspect from development to approval. That can take up to 15 years. In addition, the process is very expensive; we are now talking about about 4.8 billion euros per drug. The last thing is the uncertainty, because most drugs cannot pass the clinical trials. Particularly important here is clinical phase 2B, in which the efficacy of a drug is tested on humans for the first time. Even if the drug is neither toxic nor does not cause bad side effects and addresses the drug target, the drug may not produce a positive effect in human disease progression. In most cases, this means that the drug target that was addressed does not really have anything to do with the disease. Biotx.ai can mitigate all of these problems," Klinger explains. "We are able to find this so-called genetic support for a drug target, that is, evidence that a drug target actually has something to do with the disease. This leads precisely to the fact that not most active substances fail in phase 2B, but that the chance of success is three times higher."
In parallel, the streamlined process reduces both the time required and the overall cost of drug development. "But the ideal case is actually that we identify drug targets and then there's already a drug for them, because then we can do exactly what we're doing with Covid-19 and go straight into phase 2B, which can shorten the whole process from 15 years to 6 months."
Covid-19 treatment using revolutionary genetic predictive models
Instead of long-term vaccine or drug development, biotx.ai identifies genetic signatures that indicate the severity of a particular person's symptoms, with the goal of protecting those who will be severely affected by the disease while allowing those who are asymptomatic to return to a normal life.
"On the technical details, that's pretty medical then," Joern laughs. "We have a technology platform, that's where we look at certain genetic signatures of the human genome when the person has a disease that is particularly severe. That's what we did with Covid-19. With the help of biomarkers, such as the number of white blood cells associated with Covid-19, we came up with the drug target CDK6 (cyclin-dependent kinase 6) relatively quickly.
To understand why our therapy is so effective, you must remember that the patients who get Covid really badly don't die from the virus, but from their body's reaction to the virus. This means that the immune system overreacts and the body eats up its own cells in the lungs, which means that the patients can no longer breathe. Usually these patients are given immunosuppressants, but these downregulate the entire immune system so that immune responses that are helpful in fighting the virus are also turned off, so that a patient's chances of survival are only 50%.
The therapy that we have found is able to selectively suppress only those processes that are immune overreactions to Covid. That means the rest of the patient's immune system will be unaffected and the patient will not have this overreaction against Covid. The beauty of the whole thing is, there are already drugs for this drug target, they are used for breast cancer, so if you were to use that for Covid, you could use clinical trials in terms of toxicity and side effects. This saves a lot of time and speeds up the process from development to the patient.
The advantage over vaccines is that you can selectively target people who have really severe disease, which is particularly important when you're not able to do a long-term study of the vaccine, so potential side effects are not yet known."
A look into the future of biotx.ai
"With our current investments, we are transforming our business model from service-based to product-based. We will bring the insights from our analyses to the patient in the form of diagnostics, enabling true disease prediction. Disease prediction enables high-cost savings in the healthcare system, as novel, longer-lasting expensive drug treatments also emerge. For example, we are working with our partners on novel prediction models for Parkinson's disease and breast cancer. In the latter, early prediction can save lives without the need for expensive drug treatments. Our goal is to use novel statistical tools for genomics to contribute to better and faster treatment of patients during this global pandemic," Klinger summarizes his vision.
“We are excited to be part of the 5-HT Digital Hub Chemistry and Health ecosystem because it is the fresh outside perspective that enables startups to develop innovative solutions. And that is exactly what the changing pharmaceutical industry urgently needs.”
Become part of the 5-HT Digital Hub Chemistry & Health
Exchange ideas with innovative startups and future-oriented companies in our ecosystem. We look forward to meeting you!