‘If an animal-free research model delivers the best and fastest results, pharmaceutical companies will adopt it naturally’
MechPath brings together TNO, universities, biotech companies, and Stichting Proefdiervrij to develop an innovative approach to make drug testing faster, better, and thus cheaper – without animal testing. The researchers are working on an animal-free AI model that takes into account the differences between patient groups. This should ultimately make it possible to predict how different patients respond to a particular drug. We spoke with biochemist Roeland Hanemaaijer and data scientist Lars Verschuren, both from TNO in this collaboration.
The challenge of drug development
Ninety percent of drugs that test positively in preclinical experiments still fail during clinical studies on humans. This makes the development of new drugs a long and expensive process. Drugs that do pass clinical studies often only work for a minority of patients. ‘For liver fibrosis, for example, there is a drug that costs €40,000 per year,’ says Roeland. ‘Very expensive, and you only know after two years if it works well. It would be fantastic if we could predict for each patient whether it makes sense to start that treatment.’
Every patient is different
And precisely that prediction per patient is what MechPath – a collaboration between TNO, Galapagos, LACDR (Leiden University), and Proefdiervrij – is working towards. Starting with fibrosis, scarring in organs. Fibrosis development progresses quickly in some, slowly in others. We don’t know exactly why that is. But the fact that patients are so different partly explains why tested drugs often only work for a portion of patients. At the same time, the cell models used for testing are often based on the cells of a single patient. And that one patient can never represent all patients with their unique characteristics.
Every cell is different
Also, the variation between cells used by different research teams is often so large that research results are difficult to compare between teams. ‘Cells differ much more from each other than we used to think,’ Roeland explains. ‘Whether cells come from older or younger people, men or women, it all influences how they react to, for example, a drug.’ MechPath’s research also showed that cells remember if they are sick. ‘As far as we know, this is completely new information and quite sensational,’ says Lars. ‘Cells are isolated from a patient, frozen, sent to TNO, and cultured by us again. A week and a half later, those cells still know they are sick and produce fibrosis on their own.’
As close as possible to the patient
The holy grail in drug research is a model that is as close as possible to the patient, all collaboration partners agree on this. ‘At TNO, we are essentially between scientists doing fundamental research and companies,’ says Lars. ‘Companies have us test their drugs. Of course, they want to bring their drugs to market as quickly as possible, and to make that possible, we choose the model that best represents the patient. But ultimately, a model always represents only a small piece of a patient. To know if you have the best model, you need to know exactly what you’re measuring.’
Only after translation do research results become relevant
This may sound obvious, but according to Lars, not all researchers know their model inside out. ‘For example, there are at least five cell models on the market with which you can test the effect of cholesterol-lowering medication on liver cells. Most researchers will take the cell model that’s in their freezer, that they’ve worked with before and think they know works. But we’ve researched that which cells you use, and in what condition those cells are, is very important for the translatability of your results. And only after that translation do your research results become relevant.’
Not against animal testing, but for models that work
Results from animal experiments do not translate well to patients because animals are not humans – that’s what we always say at Proefdiervrij. And although Roeland agrees with this, he also adds a note. ‘An animal is not a human, but neither is a group of cells. Animals are living beings, you don’t want to do unpleasant things to them if it’s not necessary. So you should certainly be critical of the use of laboratory animals, but you should be just as critical of the alternatives.’
Roeland therefore prefers not to say he is against animal testing. ‘I’m in favor of models that predict well. If we manage to create animal-free models that deliver faster and better results than animal testing, pharmaceutical companies will adopt them naturally. To demonstrate this, you need to know the model, the disease mechanism, and the patient very well.’
Collaborations like MechPath are incredibly important to demonstrate that animal-free models work and to ensure that they are actually used by pharmaceutical companies. By now, MechPath’s research is almost completed. Read here what exactly they have done and achieved.