Recursion Pharmaceuticals: finding new drugs using data science

This article is a teaser for the Scottish Mortgage Recursion podcast. Recursion Pharmaceutical’s co-founder Chris Gibson tells Tom Slater why using robots to map biology could be the key to finding new drugs to treat rare diseases.

As with any investment, capital is at risk.

Discovering new medicines involves exploring the frontiers of biology and chemistry. But it’s also a ‘big numbers’ problem.

First, there is the biology problem: humans are complex beings. There are 20,000 plus genes inside each of us, encoding hundreds of thousands of proteins inside trillions of cells.

Then there is the chemistry problem: there are more treatments than could ever be tested. Researchers have estimated that there are 10⁶⁰ molecules with potential drug-like traits. For context, that’s more than 5,000 times as many atoms as our solar system contains.

In recent decades, the development of new medicines has effectively relied on researchers stumbling upon promising treatments and recognising their worth.

Acknowledging, “we are lucky if we understand 2-3 per cent of biology”, Recursion Pharmaceutical’s co-founder and chief executive Chris Gibson believes the scale of the challenge warrants a different approach.

“What we’re trying to do is leverage technology – things such as robotics, machine learning and AI [artificial intelligence] – to take a broader view and start to build maps of biology and chemistry that allow us to home in on those places where there might be a potential treatment, much more quickly and efficiently,” he explains.

Using artificial intelligence, Recursion has already mapped out more than three trillion searchable gene and compound relationships. These maps should help us to better understand biology and chemistry and how they interact, which should ultimately make it easier for scientists to navigate their way to new drugs more efficiently.

Treatments are initially developed ‘in silico’, meaning within computer software. That involves simulating how many thousands of potential drug candidates would interact with millions of disease models to identify compounds with interesting reactions.

“I tell people that we do my entire PhD’s worth of experiments every 15 minutes at Recursion, up to 2.2 million experiments a week.”

Only then do humans conduct tests in real-world laboratories to confirm or disprove the results.

“Sometimes our predictions are wrong,” acknowledges Gibson.

“But what we then done is generated new data… that can be used to retrain the model to make better predictions in that area of biology or chemistry in the future. So, it becomes essentially a learning system.”

A pharma company, but not as we know it

The process turns drug discovery on its head. Commonly, companies form a hypothesis based on existing medical literature about how to treat a disease and then hunt for a substance to match, the chemistry part of the equation. Even those employing AI typically concentrate their efforts here.

By contrast, Recursion focuses on what Gibson believes is the harder question of biology: as in studying what changes are needed within cells to treat a disease. And going back to the premise that this is a big numbers problem, it is Recursion’s ability to interrogate big data sets using machine learning that means it can test a multitude of potential drugs before forming a hypothesis.

The benefit of this approach is that it should yield speed and efficiency savings. That means the company can go after:

• rarer diseases, where patient numbers wouldn’t normally justify the large upfront costs and long timescale needed to develop new drugs
• complex diseases, where a poor understanding of the biology involved has acted as a deterrent
• diseases for which Recursion’s process has led it to form a different hypothesis to other companies

If all goes well, Recursion’s methods should lead it to new types of game-changing drugs rather than incremental improvements to existing compounds, which typify the industry.

Going big on progress

The company has come a long way in a short time since its 2013 start.

“One of the great opportunities and challenges we had in founding Recursion was that the founding team was a physician, a computational scientist and myself, a bioengineer," Gibson recalls. “And only one of us had ever worked on drug discovery as a main focus.”

They started with a “clean sheet”, he says, asking “how would you discover a medicine in a better way if you were not constrained by how it’s been done in the past”.

In answering that question, the firm has created a “new language” to help its staff share knowledge from different disciplines and technologies.

“When we bring people in, it often takes them six to 12 months to really get onboarded because there is this complexity of a new language and a new way of thinking,” Gibson explains.

Given the time involved, it's critical to hold onto those new recruits, and the firm’s goals and Salt Lake City location, away from traditional biotech hubs, help in this regard.

“Our retention levels are much higher because people feel like they're part of a mission,“ Gibson says. “And they're not walking down the street, being distracted by a dozen offers a year to go join other companies.”

Drug discovery at scale

With five drugs in clinical trials – the last stage of testing in humans before a drug comes to market – Recursion has already achieved more than most pharma companies of its size and age.

One of its most advanced programmes is a treatment for cerebral cavernous malformation. The disorder can lead to seizures, difficulty in speaking and problems with balance, among other problems.

It affects nearly five times as many patients in the US and Europe as cystic fibrosis but has received less attention to date.

“Because the biology was poorly understood, there's no companies with drugs in clinical development that we're aware of,” explains Gibson.

But the firm’s true value to its investors lies in the long-term promise of its unusual approach and technology platform.

The long-term outcomes should be new and better medicines, an improved understanding of biology and long-term growth for those who supported the company on its way.