subscribe
 

AI methods in biotechnology

20th June 2018


Work being done at BenevolentBio
Benevolent AI applies AI tech in the human health and bioscience sectors

As pharmaceutical research becomes ever more complex, artificial intelligence (AI) methods are emerging as a useful tool in the field of biotechnology and drug research – with an increasing number of biotechnology companies and research centres now using cutting-edge AI techniques for a wide range of purposes.

Blindness cure

One prominent recent example is BenevolentBio – a subsidiary of BenevolentAI, responsible for applying AI technology in the human health and bioscience sectors – which is working to find treatments and a potential cure for age-related macular degeneration (AMD), one of the leading causes of blindness. As part of this work, the company is collaborating with four UK sight loss charities to generate new insights and identify promising research areas for the condition – a decision driven by the importance of finding new treatments, particularly since, by 2020, almost 700,000 people in the UK are predicted
to have late-stage AMD, with tens of millions more across the world living with the condition.

As Jackie Hunter, CEO at BenevolentBio, explains, the AI technology created by BenevolentAI ‘generates usable knowledge from vast volumes of unstructured information in scientific papers, patents and clinical trial information, together with a large number of structured data sets.’

“It works to understand the information by employing an array of proprietary deep learning linguistic models and algorithms to analyse and understand context; then reasons, learns, explores, creates and translates what it has learnt to produce unique drug development hypotheses,” she says.

In Hunter’s view, one obvious advantage of this approach is the speed at which operations can be managed using AI – and she reveals that the company has ‘conservatively estimated’ that on a per-project basis it can cut the early stage drug discovery process by ‘up to four years against pharma industry averages.’

“That will have a big impact on the ability to bring drugs to market as we continue to evolve towards truly personalised medicine. The biggest challenge is access to data – especially negative data. Much of the negative data is unpublished and therefore it is very hard to access it,” she says.

According to Dr Laura Ferraiuolo, Lecturer in Translational Neurobiology in the Sheffield Institute of Translational Neuroscience (SITRAN) at the University of Sheffield – which continues to work closely with BenevolentBio as part of the project – it is ‘clear that the scientific community currently produces much more information and data than can be processed by a single individual, hence the need for AI to mine into large datasets and databases.’ In her perspective, AI will lead to a ‘massive acceleration’ in hypothesis generation and new findings, leading to faster drug discovery.

“Processes that now take weeks, months or years even, might take only a few hours. For what we can foresee, AI already can and will be able to perform a number of complex in silico tests, that will not only save a large amount of time, but also money, thus releasing funding for further scientific and clinical advancements. One of the current challenges is testing the hypotheses and feeding back into the system in a timely fashion, so that the machine learning process can continue and be perfected,” she says.

The Sheffield teams’ collaboration with Benevolent is ongoing, and Ferraiuolo confirms that they are currently working together to ‘understand in depth all the targets of its lead compound, and identify the exact mode of action’ – a process she expects will the team to develop a drug that has the highest efficacy and the lowest possible side effects. The team will then optimise compound dose and mode of delivery to obtain the best possible results once in a clinical trial – a process which, along with validation tests, might require ‘up to two years.’

“Once we have a strong proof of concept, we will be able to approach regulatory bodies to start moving towards clinical trials. I have no doubt that AI will be at the core of drug discovery, as we now have the capacity to mine into the immense amount of data we are able to produce. In the past 10 years we have seen a boom in the amount of chemical, genetic and biological data scientists can produce. The interesting aspect is that this data was produced to answer a specific biological question, but now we have the information, we can use that data to answer other questions and extract new information,” says Ferraiuolo.
“AI is already extensively used for image analysis in the medical field, I believe in silico models and drug discovery are our present and immediate future,” she adds.
Looking ahead throughout 2018, Hunter says that the focus of the company’s effort will be on the analysis of clinical data and genomics ‘to see how the technology can be used to impact clinical trials and patient stratification.’

“AI is going to have a massive role in personalised medicine. The power of AI to harness the available data in order to understand patient endotypes will help us more readily identify the right medicine for the right patient,” she adds.

Improving productivity


Elsewhere, Dundee-based company Exscientia is applying its AI-based drug discovery and design capabilities to dramatically improve productivity in the stage from initial chemical design to clinical candidate, the single most expensive part of drug development per launched drug.

As Andrew Hopkins, CEO at Exscientia, explains, the company’s goal is, starting from a project definition, to design and advance novel high-quality compounds with excellent therapeutic properties to clinical entry in one and a half years. It does this by applying AI to ‘dramatically reduce the number of compounds required for analysis from typically 2,500 per project using traditional approaches, to just 500 when using AI-driven techniques.’

“Our approach is to fuse the power of AI with the discovery experience of seasoned drug hunters. As a result, the company believes it is the first to automate drug design in a manner surpassing conventional approaches,” he says.

Since its establishment in 2012, Exscientia has achieved several collaborations, including more recently with Sanofi and GSK, which have helped to fuel expansion. Over time, the company has enhanced and refined AI-driven approaches, testing them in partnership with pharmaceutical partners on real projects, which also helped fund the company’s advancement without the need for external investment.

In these early partnerships, Hopkins reveals that Exscientia has ‘successfully and reproducibly completed projects to drug candidate stage within its target of 500 compounds and 12 months.’ For one partner, a candidate molecule was delivered within 12 months of project initiation. In Hopkins’ view, AI, if used correctly, most certainly has the potential to ‘transform the productivity of drug discovery and deliver superior candidates into the clinic.’

“The drug discovery phase is the most expensive phase of the drug discovery and development process – the goal is to generate better quality candidates quicker and more cost effectively. This will change the way discovery is done,” he says.

“To make it happen, Big Pharma needs to make the investment. Large companies need to look critically at their incumbent processes and the associated metrics and ask whether those are the right processes to take them towards the future of drug discovery. We are already working with (and have worked with) several pharma firms to apply our AI approaches and the results so far are encouraging,” he adds.

Ultimately, Hopkins argues that there will need to be a ‘change in big pharma mentality to such a radical rethinking’ – and he describes drug discovery as the ‘last artisan industry.’

“One of the key opportunities is also it’s challenge; the amount of data now available is beyond any human’s individual capability to hold in mind at one time. For AI focused towards drug design, where Exscientia concentrates its expertise, this provides unprecedented opportunity,” he says.

“Other companies are looking at patient stratification and personalised medicine. For example, in patient stratification for clinical trials, AI may be used to fit the patient to the treatment being tested better, enabling quicker recruitment and increasing the likelihood of getting the required clinical response for regulatory approval,” he adds.





Subscribe

 

Subscribe



Newsbrief

Twitter Icon © Setform Limited
subscribe