Can AI rescue lab rats and guinea pigs? How new technologies could solve a major ethical issue

"Drug development is currently extremely ineffective, with a staggering 89 percent of all drugs failing in expensive clinical trials, where a third of these are safety failures. And so, the inability to predict, prior to clinical trials, which drugs would work safely in the human body is a significant unmet challenge," Isaac Bentwich, CEO of Quris, told IE.

Quris –– a biotech company founded in 2020 and jointly headquartered in Boston, U.S., and Israel –– uses AI to forecast which medications would be safe to use in humans and has, in a way, revolutionized drug development.

There is no accurate data about the kinds and the number of animals sacrificed for research around the world, but the long-standing ethical questions about the use of animals for research purposes have been frequently debated in the scientific community.

Scientists are also beginning to see the limitations of using animal models as they gain a greater understanding of how human biology functions. Innovative therapeutics based on human DNA and cells or other non-traditional treatments have increased the quest for alternatives to animal testing.

"The core of our platform is machine learning [an application of AI], but we realized that AI is only as effective as the data that it is trained on," said Bentwich.

"So we use cutting-edge, miniaturized 'patients-on-chip' biology to generate data that is highly predictive of drug safety in the human body, and we use this data to train the AI."

Shift from animals to AI

Many countries and medical companies are using AI to slowly move away from animal testing, especially given the staggering percentage of failures in animal testing.

By 2035, the U.S. Environmental Protection Agency will no longer undertake or provide funding support to studies involving mammals. The measure makes the American agency the first government body to set a firm timeline for phasing-out animal research.

About 92 percent of medications that pass preclinical tests, including "pivotal" animal tests, are eventually deemed not safe and effective for use on humans, according to an estimate by the FDA (U.S. Food and Drug Administration).

More recent research indicates that the failure rate has risen and is now closer to 96 percent, despite efforts to make animal testing more predictable. Lack of effectiveness and safety issues that weren't anticipated by animal experimentation are the main reasons for failure.

Over a period of around 20 years, fewer than 10 percent of what was thought to be extremely promising drug discoveries are put into widespread clinical usage.

The failure rate of novel medications used in animal testing surpasses 95 percent, claims the People for the Ethical Treatment of Animals (PETA), a U.K. non-profit dedicated to establishing and protecting the rights of all animals.

The figures are especially clear in the case of cancer, where research has shown that 97 percent of the time that a new drug is tested in a clinical trial for a particular type of cancer, it never makes it to the market.

The industry's low productivity rate, in terms of drugs that are tested but never make it to market, demonstrates the patchy trustworthiness of animal testing since many medications that are beneficial in mice don't perform well in humans and vice versa.

Therefore, Bentwich said, Quris uses a Bio-AI platform integrated with powerful machine learning, "patients-on-a-chip" biology.

"The solution must be a 'Bio-AI' approach, which harnesses and integrates cutting-edge AI, novel engineering, and miniaturized 'patients-on-chip' biology," Bentwich said.

Roche, or Hoffmann-La Roche AG, is another big name in the pharmaceuticals and diagnostics businesses. They are a multinational healthcare corporation based in Switzerland.

"We are working heavily on leveraging modern replacement and supplementary methods and strive to use animals only where absolutely necessary," Matthias Lutolf, scientific director at Roche's Institute for Translational Bioengineering (or ITB), told IE.

The company claims it is creating cutting-edge treatments and diagnostics for serious, life-threatening, and debilitating disorders. The company says it is making significant investments in the creation of techniques for creating medications without using animals, such as computer simulations or strategies like organ on chip or organoids. These are systems containing engineered or natural miniature tissues grown inside microfluidic chips. To better mimic human physiology, the chips are designed to control cell microenvironments and maintain tissue-specific functions.

"These 'mini organs' – tiny versions of the organs in our body, e.g., livers, lungs – are grown in the lab, generated from the stem cells of individual people. They help us understand the way our organs work, how the disease develops, and to test the specific effects of medicine safely – outside the body," said Lutolf.

"The ITB aims to bridge academic and pharmaceutical research to bring the best scientists of both worlds together to lead the broad adoption of organoids in pharmaceutical R&D and clinical practice."

Modern approaches like organ-on-chip, machine learning (ML), and AI are used at Roche for analysis and have great potential in the medium term not only to replace animal testing but also to improve drug candidates or to create bespoke medications.

"After all, humans are the relevant 'biological system' for us," said Lutolf.

"Thanks to these efforts, we have been able to reduce the number of animal experiments by 40 percent in the last ten years," he said.

Patient-on-a-chip technology

The term "patient-on-a-chip" takes organ-on-chip one step further. It refers to the placement of several different miniaturized, three-dimensional organ-on-a-chip (such as the liver, brain, etc.) that are interconnected by a blood-like circulation. For example, it may include a tiny liver that is a third of a millimeter in size, a tiny 'brain,' and other miniaturized organs. This enables testing the safety of medications on these networked "patients on a chip," or miniature human organs.

"We test thousands of known drugs, on hundreds of different 'patients-on-a-chip,' and use the data that is generated to train the AI to recognize which drug is safe in the human body," said Bentwich.

Drugs that are toxic in the human body are accurately identified by Quris' platform, even some that have been overlooked by conventional biological testing and animal testing despite being recognized to be unsafe in people.

"Over the past couple of years, organ-on-chip technology has been shown to be highly predictive of drug safety, significantly more so than 'conventional' 2D biology tissue cultures," added Bentwich.

For instance, a drug could be placed on a miniature human liver, which would then metabolize it— that is, chemically break it down into metabolites — in the same way as the liver does in the human body.

The metabolites would then pass through a miniature blood-brain barrier as they would in the human body, and only then would they reach the miniature brain and interact with it.

Real-time nanosensors track how each of these miniature organs reacts to medications, and artificial intelligence is used to analyze the vast amounts of data produced.

"Quris' uniqueness is our BioAI approach, which combines the unmatched power of AI, coupled together with predictiveness of next-generation patients-on-a-chip technology," said Bentwich.

"We're now working with top pharma companies to further validate and implement this technology for their specific-use cases," he added.

Similarly, Roche's "Organ-on-a-chip" technology has been utilized to substitute animal testing for a number of years.

"It is an in vitro technology in which a human organ is simplified and miniaturized on a microchip in order to emulate and test certain biological processes," said Lutolf.

"In other words, it is an artificial replica of a physiologically interacting human tissue or organ system in a cell culture environment."

In order to reproduce some of the essential characteristics of neurodevelopmental and neurodegenerative illnesses, including Alzheimer's disease (AD) and cancer, in vitro, Roche's Lutolf said they are now investigating unique human organoid-based model systems.

"One of the grand challenges is to create functional brain tissue (with all the key cell types including neurons, astrocytes, and oligodendrocytes) that reflects the disease state of an elderly person since brain organoids have so far mimicked only relatively early stages of development," said Lutolf.

"Neurobiologists from pRED/NRD and researchers from ITB have teamed up to work on this challenge, developing innovative technologies to improve existing brain organoid models."

The company works closely with neuropathologists and brain banks all over the world and invests heavily in biomarkers to monitor disease progression or measure markers that change in blood or cerebrospinal fluid over the disease course or with therapy.

"Better understanding of what happens in humans will ultimately help us make better human model systems and reduce the need for animal models," said Lutolf.

Machine Learning (ML)

Approaches like ML play a critical role in this effort.

Lutolf said the convergence of high-dimensional data and advanced computational methods, such as ML, has the potential to elevate "our approach to drug discovery and development and transform how we bring potential breakthroughs to patients."

Roche also plans to analyze immunological control in cancer, he added.

For this research, the company is interested in creating engineered versions of human tumors that may be used as a stand-in for real cancers in order to research cancer biology (including immunology), find novel targets, and test potential new medications.

"The systems would be reductive and modular, allowing us to add or remove components at will, defining their individual and combined contributions to tumor behavior and response to drugs. The components to be studied and manipulated can, of course, be various types of immune cells from the tumor environment," said Lutolf.

"The fact that these models are built using components that were obtained from actual patients guarantees that they reflect not just patient-specific biology but also variations in tumor form, behavior, and medication response across different patient groups."

Science sees a huge potential in AI-backed technologies to completely revolutionize the way we discover and develop new medicines.

The technologies discussed here are just some examples of how science is moving away from the use of lab animals and opening the door to creating precise or tailored medicines or to using them clinically in a way that maximizes patient benefit and eliminates the need for animal testing.