AI approaching inflection point in drug discovery

It’s often said that the average cost of developing a single drug is $2.6bn (£2.2bn). The eye-watering price tag can largely be explained by high failure rates – only one out of 10 drug candidates successfully passing clinical trials and making it through to regulatory approval. Trials themselves tend to be larger and more complex than they were a few decades ago. This is where Alphabet (US:GOOG)-owned Isomorphic Labs thinks that its machine learning platform can make a transformative difference. 

The company – which was spun out of DeepMind, Alphabet’s London-based artificial intelligence (AI) subsidiary, in November 2021 – is said to be moving closer to securing its first commercial deal. Two unnamed sources cited in the Financial Times claimed that Isomorphic Labs was in partnership talks with several big pharmaceutical companies. It’s expected that a formal agreement will be announced in the coming months.

Data, data everywhere 

There is now more patient data available to researchers and clinicians than ever before. But making sense of all this input is laborious at best – which is why the idea of outsourcing analysis to an AI platform is so appealing. To make a new drug, scientists typically screen millions of molecules before zeroing in on a handful that may have the desired therapeutic properties. Numerous rounds of testing follow before a drug candidate even makes it into the earliest clinical trial rounds. 

AI has the potential to analyse data gathered at every stage of the drug discovery and development process. Analysts with Morgan Stanley’s research division estimate that even modest improvements in early-stage success rates could result in the creation of 50 novel therapies over a 10-year period. According to a note issued in November, this could also translate into a new market worth some $50bn for the companies that make these drugs. 

Although the sector’s largest players have been interested in using AI in the research and development process for some years, early experiments were not a resounding success. In 2016, Pfizer (US:PFE) announced it would be trialling a drug discovery programme with IBM (US:IBM) machine learning tool Watson.  

As novel as this might have sounded, the tool ultimately didn’t revolutionise pharma R&D and IBM stopped selling it in 2019. However, investors clearly continue to believe in the technology’s potential – Boston Consulting Group found that third-party investment in AI-enabled drug discovery more than doubled annually in the five years to 2021 to reach $5.2bn. 

Possible inflection point?

Last year’s tech funding environment was notoriously difficult, and it’s likely that enthusiasm for machine learning in pharma will have wavered in accordance with that shift. Morgan Stanley research analysts are nonetheless predicting that the sector will reach an “inflection point” in the next two years driven in part by increased collaboration between AI technology developers and large pharmaceutical companies.

In 2019, Astrazeneca (AZN) partnered with BenevolentAI (NL:BAI), an AI-focused drug discovery and development company, to identify “novel targets”, or therapies entirely new to medicine. As of October, five potential drugs had emerged from the collaboration – two for chronic kidney disease and three for idiopathic pulmonary fibrosis, a lung condition. According to the London-headquartered BenevolentAI, its drug discovery platform is “disease agnostic” (meaning it can be applied to a multitude of illnesses) and capable of rapidly generating novel drugs at scale. 

The UK’s other blue-chip pharmaceutical company, GSK (GSK), was even quicker to embrace AI, signing a partnership agreement with Exscientia (US:EXAI) in 2017. Like BenevolentAI, the Oxford-based firm has also developed a drug discovery platform that utilises machine learning technology to identify novel compounds. Two years later, the duo announced they had jointly created a drug candidate for chronic obstructive pulmonary disease, although the testing timeline remains unclear. And Exscientia’s share price drop of 75 per cent in 2022 shows AI companies were far from  immune to the biotech sell-off. 

At the same time, GSK has built up an in-house AI team of more than 120 engineers, currently the largest in the industry, while Exscientia has also signed a collaboration agreement with France’s Sanofi (FR:SAN). Other pharma firms with dedicated AI teams include Novartis (US:NVS), Roche (CH:ROG) and Bayer (DE:BAYN). 

As AI in drug discovery matures, Morgan Stanley analysts say investors “will have to weigh how individual firms are using AI and machine learning to develop drugs”. While they might be investing heavily in the technology, it’s not yet clear which companies, partnerships or platforms have a genuine advantage in drug development. The highly technical nature of the space and secretive competition between the pharma giants means investors are unlikely to be able to build up their own expertise in the space, so clinical trial results should be watched closely. 

“If initial read-outs are consecutively strong, we believe stocks across the space could rise as investors gain confidence in a well-defined total addressable market for AI-enabled drug development,” said Vikram Purohit at Morgan Stanley.