We've updated our Privacy Policy to make it clearer how we use your personal data. We use cookies to provide you with a better experience. You can read our Cookie Policy here.


AI in Pharma – Adoption, Disruption and Drug Discovery

Listen with
Register for free to listen to this article
Thank you. Listen to this article using the player above.

Want to listen to this article for FREE?

Complete the form below to unlock access to ALL audio articles.

Read time: 4 minutes

SLAS2020 began with an in-depth look at how artificial intelligence (AI) is changing the landscape of drug discovery. On January 27, hundreds of attendees welcomed Jackie Hunter, Ph.D. to the stage as she offered those listening key insights gained from over 30 years in the bioscience research sector and knowledge gained as BenevolentAI’s board director.

Describing AI as her “favorite subject” Hunter began by emphasizing the power of AI in numerous industries – automotive, financial, agricultural – and pharmaceutical.

“AI is extremely important in healthcare – and the pharmaceutical industry in particular. One of the reasons for this is the digitization of human health and human healthcare – and the incredible “ramping up” of data sets,” said Hunter.

She explained that healthcare institutions worldwide are dealing with an almost 10-fold increase in data per year since 2016 – clearly highlighting the need for ways to process and manage the ever-increasing volume of data emerging.

With such a vast quantity of data being generated, the healthcare sector and pharmaceutical industry must be able to harness data more effectively to enable the discovery of new medicines and to deliver healthcare to patients more efficiently.

That’s where the power of machines comes in to play. According to Hunter: “The only way to do this is to use AI.”

Taking a closer look at AI

Artificial intelligence is a term that describes more than just a single “thing”. AI can be subdivided into different types. Hunter defined these types as; machine learning, deep learning, natural language, robotics and visual analytics.

Hunter described machine learning in detail, defining unsupervised machine learning as the: “automation of analytical model building”, whereby the machine is provided with data that it analyzes. It subsequently identifies unknown patterns in that data. “You can build on that model and make decisions or predictions without the need for human intervention,” explained Hunter.

“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” – Roy Amara.
“Supervised machine learning is when the machine is given a set of sample data and training data.” This type of learning enables the generation of a predictive model that is based on both input and output data – in contrast to unsupervised which is solely based on input data. “The sample data includes the outcome you are looking for – for example a patient with disease X. The machine can then learn from that data and it is then given a set of data it hasn’t seen before.”

Deep learning – a subset of machine learning – can be thought of as “learning by example”. It is based on artificial neural networks that can learn from unstructured or unlabeled data. “These neural networks are able to extract high-level abstract features from very large complex heterogenous datasets – this type of dataset and this type of intelligence is very appropriate in the context of biological analytics.”

Whilst Hunter refrained from a deep discussion into natural language processing, robotics and analytics – she did clearly emphasize their importance in the context of drug discovery and pharmaceutical development, explaining that there had been: “huge leaps and bounds in terms of visual analytics.”

AI in healthcare – who’s already reaping the benefits?

Hunter explained that whilst several areas are benefiting from AI, including biomarker development and drug discovery, the two “key players” were currently pathology and radiology.

When it comes to pathology, numerous studies have demonstrated the ability of AI systems to provide accurate diagnosis and treatment decisions (comparable to that of expert pathologists) for cancers including; prostate, breast, and brain.

On May 08, 2019 the US Food and Drug Administration (FDA) granted "Breakthrough Device" status to a company developing a retinal AI-imaging platform that is able to analyze eye scans for biomarkers linked to neurodegenerative diseases such as Alzheimer's disease. Last year researchers developed a novel AI-based tool to predict risk of breast cancer – the deep learning model yielded substantially improved risk discrimination over the current clinical standard (Tyrer-Cuzick model). And most recently, in January 2020, the FDA cleared the world’s first radiology AI solution to aid diagnosis of stroke.

Adopting AI in pharma – why is there a need?

Advances in robotics have enabled drug discovery teams to screen an unparalleled number of compounds at a much higher throughput than ever before. Lab automation developments coupled with the evolution of in vitro cell screening, whereby researchers are now using three dimensional models of unprecedented complexity, has fueled the need to adopt “more intelligent” bioinformatics and AI solutions to manage and analyze the data generated.

“It’s all very well having the technology, but you also need to have the tech-readiness – of the system, the organization, and the people,” said Hunter.

“The pharmaceutical industry, as it is currently configured just isn’t sustainable.” – Jackie Hunter, BenevolentAI.
With the staggeringly high rate of drug development failures, there is a clear need to find ways to increase success and decrease the cost of bringing a drug to market. In a nutshell, AI will lead to quicker, cheaper, and more effective drug discovery. AI coupled with robotics will help tackle issues with reproducibility and AI will help teams make better decisions on which therapeutic targets to prioritize.

Virtual screening (VS) is a great example of how powerful AI can be. VS involves interrogating vast compound libraries in silico. This approach can help expedite the development of a drug – a process which is typically extremely costly and has a high rate of attrition. AI methods help to predict which compounds will be “most favorable” in terms of binding to the therapeutic target.

Implementing AI – what’s the issue?

So, you might be left wondering “What’s the hold up? Why isn’t every drug discovery company implementing AI?”. The answer – there are numerous “obstacles to swerve”, and challenges linked to the adoption of AI approaches.

“I certainly believe that AI offers great potential,” – Jackie Hunter, BenevolentAI.

Hunter described the recruitment and retention of talent as a key challenge related to the implementation of AI, highlighting that: “Data scientists can go anywhere.” Having ways of incentivizing the different skill sets that are required is key to successfully harnessing AI. You must also ensure that you have people on board that are willing to embrace the technology. You may have an experienced chemist – but there may perhaps be a natural level of subconscious bias. A system, on the other hand, is unbiased. The medicinal chemists must understand the value of AI and should be willing to work collaboratively with machines to cocreate.

“You need a true marriage of the two – where both the machine can be challenged, and the chemist can also be challenged,” says Hunter. “This is what we really try to do at BenevolentAI.”

Hunter also underlined the value of diversity: “I want to highlight the importance of cross functional working – having cross functional teams. There must diversity of thought and diversity of discipline.”

Hunter advises that more experienced scientists in leadership positions should be tapping into the millennial’s ways of thinking – these are the next generation of data scientists after all.

Hunter’s take-home message: “This technology will be transformational.”