AI + data = drug discovery

December 2019

Much has been predicted about the impact of artificial intelligence (AI) on just about every aspect of life. What impact should we expect to see on life sciences research specifically? This already seems clear.

Take for example, the announcement in April 2019 that AstraZeneca and BenevolentAI have formed a long-term collaboration to use AI (essentially sophisticated algorithms and machine learning) for the discovery and development of treatments for chronic kidney disease (CKD) and idiopathic pulmonary fibrosis (IPF). And, then again, in September, Benevolent AI teaming up with Novartis's precision medicine team to investigate further their existing oncology pipeline.

These diseases are complex and the underlying disease biology is not clearly understood. Research requires the interrogation of vast datasets. Both collaborations are therefore intended to combine known biomedical, genomic, clinical and molecular data, which has been extracted and contextualised using BenevolentAI's proprietary 'knowledge graph', with a target identification platform.

This uses relation inference models to help predict potential non-obvious disease targets. They then generate potential drug-like molecules with optimised treatment properties, for synthesis. In other words, computational techniques are being used to identify biological features, or 'biomarkers', that suggest certain drugs can be used for particular indications and/or subsets of patients within a particular disease group, and working out what those drugs are.

The use of AI in combination with large datasets therefore promises to drive new and more targeted treatments. It is a trend that is not going to be reversed. Indeed, BenevolentAI claim that 90% of the world's data was produced in the last two years alone.

Right on trend, NHS England also has plans to use the wealth of data that it has accumulated over 70 years, once depersonalised, to identify groups of people who are vulnerable to health risks and predict which individuals are likely to benefit from healthcare interventions. The NHS also expects this data to be beneficial to companies within the life sciences sphere, as it makes it available to industry with the aim of driving research and innovation.

Moreover, Matt Hancock, the Health Secretary, has the ambition to launch a "genomic revolution" in which all children will be offered genome sequencing at birth, as a matter of routine. This could allow parents the choice to be alerted to specific disease risks, with the potential for the NHS to offer more personalised treatments: "Predictive, preventative, personalised healthcare – that is the future of the NHS – and whole genome sequencing and genomics is going to play a huge part in that."

Then there is what is being claimed as the world's largest genome sequencing project. Announced on 12 September, a consortium of UK Research and Innovation (a UK government agency), The Wellcome Trust, and four pharmaceutical companies (Amgen, AstraZeneca, GSK and Johnson & Johnson) is to fund the genome-sequencing of 500,000 individuals at UK Biobank in Manchester.

Such is the speed at which an individual's DNA can be sequenced that the results of this work are expected in Spring 2020, after which the four companies investing in it will have a nine month period of access before the data is opened up for public access.

It should not be forgotten, however, that there is a lot of work and risk that lies between identification of a drug target and bringing a new treatment to market. But, the ambition is that data will boost the possibility of finding the root causes of disease in specific, and smaller, groups rather than addressing merely the symptoms of whole populations.

It is also hoped that, asthe 'in-silico' methods offered by Benevolent AI and others, will increasingly replace the costly and time-consuming traditional 'wet-lab' techniques for drug discovery, the costs of drugs will come down. This is important, because one side-effect of providing treatments for ever smaller groups of patients is that the price of treatment increases per individual in order to cover the high research cost. This has already become a significant political issue.

If you have any questions on this article or would like to propose a subject to be addressed by Synapse please contact us.


Paul England

Paul is a senior professional support lawyer in our London office, specialising in patents law.

"AI in combination with large datasets promises to drive new and more targeted treatments. It is a trend that is not going to be reversed."