Andrew Hopkins, founder of pharmaceutical startup Exscientia, believes artificial intelligence will help discover new treatments against diseases including Covid

It is easy to forget, now that it is synonymous with the first Covid vaccine, that Pfizer used to be most famous for the little blue pill. “I joined Pfizer the week Viagra was launched,” says Andrew Hopkins. “It was crazy.”

Hopkins’s own work has always involved attempting to discover new drugs. As a PhD student in Oxford, he tried to find molecules to combat HIV, for which only one treatment, AZT, then existed.

It was painstaking work with long hours. “Lab meetings typically took place at midnight,” he recalls. One night, walking home at 2am, he had a brainwave. “I thought there has to be a better way to design drugs, a way of making it so simple that an individual could do it rather than a huge team.”

A quarter of a century on, the need to accelerate, improve and simplify drug discovery has never been more pressing. Just 1 in 1,000 molecules identified as potential treatments by pharma companies makes it as far as clinical trials in humans – trials that then take an average of five years and eliminate another 90pc of chemical candidates. For each drug that does make it to market, usually more than a decade after its key molecule was first identified, an average of $2.16bn has been spent on R&D.

Hospitals have become better at treating patients with Covid-19 as more repurposed drugs are approved to help fight the disease

Credit: ABED AL HASHLAMOUN/EPA-EFE/Shutterstock

Nothing better illustrates the cost of such a delay than the current pandemic. As a result, the Covid treatments (as opposed to vaccines) in use at present involve the repurposing of existing drugs (such as the anti-inflammatories tocilizumab and sarilumab, just approved for use in the UK) rather than the administering of new pills specifically designed to tackle the virus.

A pharmaceutical revolution

Today, however, Hopkins’s doctoral dream may at last be coming true. As the 49-year-old chief executive of start-up Exscientia, he is at the forefront of a pharmaceutical revolution poised finally to upend the laborious, time-consuming, expensive process – riddled with frustration and failure – of drug discovery.

At its heart is artificial intelligence. Traditionally, says Hopkins, drug discovery has been a “liberal science” – a blend of “chemistry, physics, computer science, data informatics, and all types of biology” – requiring input from specialists in very different disciplines, as well as a healthy dose of intuition and luck.

At Exscientia, which is valued at around £180m after a £49m funding round last year (and in which Hopkins holds a 40pc stake), computers trawl through databanks of chemical molecules, examining the qualities of each to identify those that might be useful against illness and disease.

Like a vast wave, this idea is washing through hundreds of companies in the industry, from established pharma giants down, all bewitched by the thought that it could lead not only to the discovery of new drugs, but also do so quickly and cheaply.

World first

Last year, Exscientia became the first in the world to have such an AI-designed molecule submitted for human trials. DSP-1181, as it is known, is a potential regulator of serotonin, a chemical in the brain that is believed to play a role in Obsessive Compulsive Disorder (OCD). Instead of the typical five years, it took only 12 months to get DSP-1181 to trial.

“We are letting the algorithms generate the compounds… fundamentally changing the relationship between what we call the designers, the chemists, and the technology,” says Hopkins. Humans, he insists, are still vital, and the company’s AI platform is called Centaur, after that mythical mix of man and horse, as if to prove it. This time, though, the augmentation of mankind’s powers comes through the addition of silicon, not hooves.

In a couple of weeks’ time, Hopkins reveals, Exscientia will start “the clinical trials of the second [AI] drug”. He won’t reveal its target, but admits that it too, had a discovery time cut from the standard of four to five years to between eight to fourteen months.

Exscientia will soon start clinical trials of an AI drug

Credit: FatCamera

Such shortcuts are possible because Centaur helps narrow down the field of potential winners. Rather than testing compounds based on many hundreds or thousands of molecules, Exscientia’s algorithms help it zero in on those likely to be successful. “We’ve actually discovered a compound within fewer than 200 molecules.”

Such acceleration also explains why Exscientia is one of 37 companies and institutions, including pharmaceutical Goliaths, to which Hopkins’ company is now playing David, selected for the Corona Accelerated R&D Europe plan – a consortium hunting Covid treatments that is the biggest of its type, ever.

A new type of drugs

“We’ve already got novel chemistry, new designs, new compounds against one of the key enzymes in Covid,” says Hopkins. But he reveals an ambition that goes much further – to develop drugs that work against all coronaviruses. “We’ve had three coronavirus pandemics in the past 20 years,” he adds, referring to SARS, MERS and now Covid. “What we require now are novel, pan-coronavirus agents, because even a small mutation can sometimes make vaccines inadmissible. We need a new armoury for the next coronavirus that might come along.”

It has all taken far longer than he hoped. Born outside Neath, in Wales, he saw for himself “the real creative destruction of capitalism and economics and technology” during a stint at Port Talbot steel works, where productivity soared even as worker numbers crumbled. Somehow, however, the pharma industry seemed immune from “the same Darwinian winds of change”.

In 2009, two years after he left Pfizer, as a professor of medicinal informatics at Dundee University, came a pivotal period when he helped show that medical data about the efficacy of drugs on proteins in the body, until then largely on the yellowing pages of paper journals, could be digitised and used in reverse – effectively to predict which kinds of other molecules might work as treatments. A year later, he founded Exscientia.

There is still a huge amount of that old paper information to mine and sift through. “The ideas to how we cure Alzheimer’s and some of the big diseases could already be with us, published in academia,” he says. “It’s about how we explore those risky ideas in a much more cost effective manner.”

Tipping point

Now his plan is to expand from the business of purely predicting effective molecules to the task of predicting drug toxicity in humans, which can often surprise researchers and halt trials apparently out of the blue.

“Everything can be improved with AI,” says Hopkins, including trials themselves and manufacturing. The whole process from molecule discovery to sticking foil on the blister pack of pills can and should, he suggests, be automated. “The company of the future is a true AI-enabled, end-to-end pharma tech. AI is one of those generational revolutions in medicine that opens the possibility for new types of company to be created.”

His excitement is palpable. Only now, he says, has his industry reached “the tipping point” where computers and algorithms are powerful enough to navigate and direct the infinite complexities of our biological systems.

That point crossed, however, he is sure there will be no looking back. “We are the only company that has an AI designed drug in the clinic,” he notes. “[But] by the end of this decade, I expect all drugs entering the clinic to be designed with AI.”