Made in Israel: First AI-designed antibody could lead to eradication of tumors

In recent weeks certain doctors and patients with terminal cancer in Australia have been participating in a highly important experiment. The doctors are injecting the patients with an antibody that they hope will activate a molecule familiarly known as IL-2, which is naturally produced in the human body and can eradicate tumors.

What makes the experiment unusual is that the antibody they're injecting wasn’t produced by living tissue, but rather by computers in the laboratory of Biolojic Design in Rehovot. The antibody, known as AU-007, is the first to be designed by computer and reach the stage of clinical trials. It evokes keen hopes because if it works, it paves the way for the development of a new kind of drug based on computational biology and big data.

Like practically every drug that enters clinical trials on humans, Biolojic Design's antibody was first tested on mice. All evinced positive reactions to the treatment. In the 17-day trial period of the study, the antibody led to the complete elimination of the tumors in ten of 19 mice, and significantly inhibited the development of tumors in the nine other mice.

Prof. Yanay Ofran, founder and CEO of Biolojic Design, is keeping his enthusiasm strictly curbed. "We have a joke we tell at conferences. 'We have great news for all the mice in the audience. We've managed to infect and sicken them with 1001 diseases and cure them.' The lingua franca of the drug development world, the empiric language it uses, is animal studies. You have to show success with an animal trial or you won't be able to raise money, the regulator won't let you test it on people, and doctors won't refer their patients to the trial."

But, Ofran explains, animal trials are not some sort of truth test that spots the good drugs and eliminates the risky ones. "I'm pretty sure there are a lot of people who paid for this [clinical trials on animals] with their lives," he says.

He isn't anti-vivisectionist, Ofran stresses. Twenty years ago he wrote a controversial column in Haaretz arguing that confining animal experimentation to life-saving purposes would destroy scientific research. He did then and does now believe animal experimentation to be a crucial tool. But when it comes to drug development, if anything he believes the method just causes confusion and damage.

Taking an example from his world, Ofran notes immunotherapeutic drugs that encourage the immune system to attack tumors, such as pembrolizumab (Keytruda). When they emerged in 2011 they proved able to dramatically extend the lifetime of many metastatic melanoma patients, who would previously die within months – and even cure about a fifth of them. Unfortunately, still, most patients are not cured; and a majority don't even respond to the treatment.

"In the case of immunotherapeutic drugs, you can have brothers with the same cancer who get the same therapy, and one heals while the other doesn't respond at all," Ofran says. "One develops no side effects and the other develops side effects that almost kill him. And these are brothers whose father is sitting waiting for them in the corridor. What do they tell us about animal trials? They say try the drug on mice, whose common ancestor with us died out 60 million years ago, and if it doesn't work on mice, or is toxic to them, there's no point in trying it on people. You know how many drug companies came to me and said, hey we have a drug we think is great but it doesn't work on monkeys. Can you make it work on monkeys even if it slightly compromises its activity in humans?"

Who Killed George Washington

Ofran, 49, a professor of computational and structural biology at Bar-Ilan University, is an impressive, eloquent person who turns science into accessible stories. In 2007 he co-hosted the documentary series (Hebrew), "Did Herzl Really Say That?" which engaged in social and cultural issues in Israel and was broadcast on Channel 8.

He opens his introductory course to statistics with the story of U.S. president George Washington who was effectively done to death by his doctors in 1799. These dignitaries wrote in their diaries about their frustration that their blood-letting efforts, in response to a cold Washington came down with, did not improve his state and that his blood turned thick.

A reporter reading the journals wrote that blood-letting was an utterly moronic procedure and accused his doctors of killing the former president. They sued him for libel and he lost after the medical establishment took the doctors' side.

"The only way we can know if something really works is statistics, not theory," Ofran tells his students. "It has to be tested on a lot of people. Theory said blood-letting should work and only when it was tested using statistical clinical tools, did that turn out to be untrue."

Ofran grew up in Jerusalem and attended a religious school. In 12th grade, senior year, he participated in an Education Ministry exchange program between schools in different cities ("Shminiyot L'Ayarot Pituach"). He moved to Acre, studied at the kibbutz yeshiva in Ein Tzurim, served in intelligence forces in the army and did a BSc in biology and physics.

He also gained broad knowledge of philosophy and linguistics. He began, and stopped, doctoral studies twice, in linguistics and brain research, after deciding that the questions in these areas are more interesting than the answers. By age 28, however, he had begun and completed a doctorate, in molecular biophysics, at Columbia University in New York.

Ofran is the third of seven children. His siblings engage in medicine, psychology and education. His grandmother used to say, he recalls, that her grandchildren had chosen their professions poorly. "A shoe salesman doesn't wake up in the middle of the night and say, 'Was 36 really the right size for that customer?'"

His father, Abraham Ofran (Zaafrani) immigrated to Israel from Morocco as a boy. He worked in organizational and methodology examinations at Bank Mizrahi Tefahot ("what he cared about was how to do things more efficiently").

His mother, Dr. Mira Ofran, is a physicist specializing in teaching physics and mathematics to children. Presently she is editing and preserving the writings of her father, the scientist and thinker Yeshayahu Leibowitz. His grandmother on his mother's side, Gerta, was one of the first women in Europe to pursue a doctorate in mathematics and worked as the scientific secretary of the Hebrew encyclopedia.

Ofran has four children aged 11 to 18. His wife, Rana Samuels, did a doctorate on using computers in climate change science, specifically looking at how climate changes affect water resources. They met in their early 20s, when Samuels came to live in Israel for a year. She died of cancer in 2014. Biolojic Design was established a few years earlier.

Powder and Gloves

Biolojic Design's story began in late 2008. Ofran, then 35, had already designed molecules that bind to one another a decade before and wanted to advance to drug development. His rationale was to "take everything biology knows about the molecules involved in a specific disease, identify which of them to attack in order to fix the problem, and to use computer power to design an accurate antibody with minimal side effects."

In other words, to design a drug that is itself a biological antibody, the kind produced naturally by the body against invaders.

But the chance of raising large sums in academia is remote. He decided the best option was to forge collaborations with Big Pharma. "Drug companies have tools to test our solutions. They have the knowhow in organizing clinical trials and every one of the big pharma giants has more resources than all of Israel's universities together," he explains.

The company embarked on its journey in 2010 and for some years sold technological services to drug companies such as AstraZeneca, Novartis, Nektar Therapeutics and Eli Lilly. It still works with some of them. These collaborations brought in a few tens of millions of dollars in income over the years, keeping the company balanced with the help of modest sums, in the terms of biotech companies, raised from investors.

Since its establishment Biolojic has raised about $30 million from investors including Marius Nacht's fund aMoon, Sami Segol’s Keter Plastics, and private investors including Prof. Chezy Barenholz. It also raised $40 million from the AppleTree Partners venture capital fund, with which Biolojic set up a subsidiary named Aulos Bioscience to promote AU-007 to a clinical proof of feasibility.

To date, though the company provided the requested molecules to the drug companies with which it worked, none of these commissioned drugs reached the stage of human clinical trials. The protracted process and heavy costs involved lead drug companies to prefer solutions that are similar to ones that worked in the past. "The companies pay peanuts for the development and promise future royalties from the drug, when it starts selling.

It could be billions, but in 15 years, by which time thousands of explanations arise why the drug – even if it's a good one – gets stuck in the organization," Ofran says. "To promote a drug you need to understand not only science and technology but organizational politics. Today we only continue collaborations when we are confident the top people at the company are committed to the project. Take Eli Lilly: We're working on a diabetes drug with it."

When choosing a drug to develop independently, Biolojic is guided by four questions: Is the biology understood, is there an unmet need, whether their technology can solve the problem and whether the drug can be brought to human trials in a reasonable time at a reasonable cost.

All four have to be answered in the affirmative. But if the last one is not – the cost is too high, or the time too long – it may be suitable for collaboration with a major drug company.

"Biotechnology companies usually say, 'we attacked a certain biological mechanism with drugs and it didn't cure the disease, so let's look for another mechanism we can attack with the same kind of drugs and maybe that will work.' They're searching for a new biology and trying to treat it using old technology," Ofran says. "I say the opposite: Let's take the existing, well-understood mechanism, the one with which we tried to attack the disease and failed, and see if we can attack it with smarter technology. Our innovation is technology, not biology."

The AU-007 antibody was designed based on the vast amount of knowledge built up on IL-2, which can act against tumors but which also has wicked side effects. IL-2 may inhibit the immune system, or have toxic effects leading to vascular leakage or pulmonary edema.

Designing by computer enabled Biolojic to produce an antibody that binds to IL-2, leading it to activate the immune system's against the tumor, but without the toxic effects. The traditional method of antibody production does not allow such fine control of their activity. Many antibodies presently used in treatment were produced in mice which had been injected with a protein associated with cancer.

Biolojic’s AI platform mimics the human immune system and instructs the computers to design antibodies based on examples and statistics. To simplify, an antibody is like a glove tailored to the molecule we want to attack, Ofran explains.

“We show the computer a million hands and a million gloves that fit them. Eventually, we showed it a hand for which there is no glove, and tasked the algorithm with creating a precise glove for it."

Before the computer finds the perfect glove, it gives the Biolojic engineers millions of "drafts", which they send to a DNA synthesis company. That company returns an envelope to Biolojic, containing tiny vials of DNA dissolved in solution. These DNA molecules are inserted into yeast cells in large test-tubes.

The yeast cells "read" the DNA sequences and produce the "draft" antibodies. Biolojic scientists then test these antibodies using a cell-sorting machine that looks like a large printer.

These machines run billions of measurements overnight, testing the interaction between “draft” antibodies and the target they should attach to identify the best one. The "printer" automatically sends the results of these measurements to Biolojic’s computers, where the AI algorithms analyze them and propose a new antibody that will do the job.

"The Broom"

Ofran takes a keen interest in politics, economics, and technology. He says that the three players in the pharma world – the regulator, investors, and science – know the industry is stuck fast and blame each other. Meanwhile, because of the way clinical trials are structured under the law, we're missing out on important developments, he suspects.

For instance, a drug might only be approved for human trials in terminally ill patients, which means the treatments accessible to the early cancer patients are only ones that proved efficacious in people with advanced disease. He also notes the example of Biogen's Alzheimer's drug. Since its approval a year ago it has been a commercial flop, Medicare elected to limit coverage of it, and this month, the company's CEO resigned.

The results of Biogen's clinical trials were not amazing, Ofran says: "They showed improvement in brain tissue but no improvement in memory. In other words, the drug did do something that seemed good for the brain, but apparently treatment needs to start earlier. Biogen asked for approval based on these results, intending to show, over time, that it also works in earlier-stage patients.

The drug was approved by a margin of a single vote by the Food and Drug Administration, the first to be approved for Alzheimer's in almost 20 years, but the medical community balked: ‘You haven't shown that it really helps, so we won't prescribe it’" – and the insurance companies of course shied from it too.

"I wonder where the public interest is," Ofran says. "Is it to make sure such things never happen again? Or au contraire, that a chance should be given? There are no easy answers, and obviously if you are on the side of a multi-billion-dollar company, then public sentiment is automatically against you."

The decision to start the clinical trial in Australia had been strategic, he admits. The various regulators have essentially different approaches. The American watchdog says "Prove the probability of harm to the patient is negligible," while the Australian says, "Prove that it's safe and the probability of helping patients is high".

"The Australian regulator provides access to patients who can't be helped any more, giving them hope even if it places them at some risk," Ofran explains. "This means, among other things, that the probability of a drug helping the initial participants in U.S. clinical trials is low. The permissible dosages are tiny: the regulator is saying 'Take it easy, when you're absolutely sure it won't harm ten patients we will consider if we can help the 11th and 12th."

Lately, the company has been working on smart antibodies capable of measuring substances in the body and reacting to different concentrations of these substances. Meaning, drugs that have more than a single working plan and can adapt their activity to different tissues, different situations and different patients.

One such drug can locate chemotherapy molecules in the body and lead it to the tumor site. "Today when you give chemotherapy to a patient, it disperses uniformly throughout the body, which is why patients lose their hair, their intestinal functions, and their nerves die. Chemo hits every site where cells are dividing," Ofran explains. "That's why chemo can't be given in higher dosages because the patient will die. If the chemo could be concentrated where the cancer cells are, the patients could be given doses that kill effectively the tumor with fewer side effects."

In the company they have dubbed this drug “The Broom”, because it sweeps chemotherapy molecules from the body's periphery and concentrates them where the tumor is. The antibody was designed to identify the chemotherapy molecule as well as the molecules characteristic of a cancerous environment so it can bind to both of them.

When the antibody identifies a chemo molecule in a non-cancerous environment, it captures it, preventing it from doing harm; then, when it reaches a cancerous environment it releases the chemotherapy in order to bind the tumor-specific molecule. Moreover, if the antibody is bereft of chemo molecules but reaches a tumor, it behaves like a regular anti-cancer antibody that attacks the tumor-specific proteins. The company hopes to bring this drug to human clinical trials in about two years.