In recent decades a strange confluence has formed in the pharmaceutical industry. A range of powerful technologies have brought enormous new capabilities to the sector, productivity has been simultaneously declining. Deloitte's Centre for Health Solutions estimates that projected returns on investment in pharma R&D have fallen to 1.9 percent, the lowest level since 2010.

The paradox of improving technology yielding deteriorating results can partly be explained by the cognitive limitations of the humans who deploy these complex tools and techniques.

© Exscientia
© Exscientia

Exscientia believes it can unravel the contradiction through automation.

The Dundee-based startup has developed a drug-discovery system that use artificial intelligence algorithms to find new candidate molecules for medicines.

"The invention of the molecules is done by the algorithm itself," Andrew Hopkins, CEO of Exscientia, tells Techworld. "Basically, we're going from data to a generation of IP by an algorithm. That we believe is a big advance in the tech world."

Read next: GTN discovers new drugs with machine learning and quantum physics

The company has already discovered four candidate drug molecules, which are now being extensively tested in late-stage preclinical studies. Hopkins predicts that by 2020 up to two of those molecules will be ready for human trials.

"They would represent the first molecules designed by an algorithm to enter into humans," he says. "That would be a great milestone for our field."

Origins of discovery

Exscientia was founded in 2012 as a spinout from the University of Dundee, where Hopkins is the chair of Medicinal Informatics.

At the time, there was not enough data to build its machine learning models. Members of his team helped some of the earliest databases used for this method and wrote some of the first serious machine learning studies using this data.  By the end of Exscientia's first year in operation, the team had published a pioneering a paper in Nature detailing the approach – a pivotal point in this new paradigm of drug discovery.

For most academics, that would be the end of their work on the subject. But Hopkins wanted to convert his invention into an innovation, by finding a way to apply it in the commercial sector.

He founded Exscientia to develop the science into a product, taking the name for the company from a Latin term that reflected his method.

"It actually comes from the motto of the US Navy, which is 'ex scientia tridens': from knowledge comes sea power," says Hopkins. "But really, we've more like 'ex scientia medicinae': from knowledge, drugs."

The first task of the new company was to validate the technology on real-world drug discovery examples, which led Exscientia to Osaka, Japan, where it teamed up with the Sumitomo Dainippon Pharma to develop a number of preclinical drug candidates. The results assured Hopkins that his approach had enormous potential.

Hopkins now believes that his method can cut the time taken to take a drug from an idea to a clinic from five years to 12 months. He estimates that this could reduce the cost of bringing a drug to market by 30 percent, an enormous saving given that the current sum is around $2.7 billion on average, according to figures from the Food and Drug Administration of the USA.

Read next: Meet the bioinformatics startups applying AI and machine learning to genetics to bring precision medicine to Europe

A spate of deals with pharmaceutical giants including Celgene, Sanofi, GSK and Evotec followed. One recently led to the discovery of a molecule that targets two distinct, biologically-validated pathways related to inflammation and the progression of fibrosis, which is attracting ongoing investment from French pharma company Sanofi.

This small molecule could in time be taken as a pill that hits two drug targets, an unusual method in the market, but one that can increase the efficacy of treating complex diseases.

"In this particular project, I believe we evolved nearly 100 billion novel compounds to identify the chemical series," says Hopkins. "It's not just a case of looking for a needle in a haystack; it's more like looking for a needle in an entire farm, to find the exact chemical solutions to this particular problem."

Bringing science to market

AI drug discovery is attracting increasing interest from big pharma, with the UK's BenevolentAI and GTN among Exscientia's leading global competitors for the investment. Hopkins says that his company stands out from its rivals as it was the first to use this method and has since established a record of delivery that the others can't match.

Read next: BenevolentAI Q&A: Techworld meets Dr Jackie Hunter, CEO of BenevolentBio

That record will show its true value when the molecules are developed into drugs on the market that can effectively treat illnesses. Hopkins admits that this will take time to happen, and is focusing on the preclinical and early-stage drug discovery phases for the next couple of years, but the ultimate objective is to help the general public.

"Many of the potential new ideas for curing Alzheimer's and diabetes probably exist already within academia," he says. "We think that by lowering these barriers of entry that you can bring ideas forward to the clinic faster and hopefully bring more innovative treatments quicker to the patients."