Artificial intelligence in drug discovery

Artificial intelligence (AI) is the ability of a machine to exhibit human-like intelligence – writes Eliott Lumb.

Computer science and large, robust datasets develop algorithms that mimic the cognitive functions of humans; they learn, solve problems, identify patterns and generate predictions. 

 ‘Narrow’ AI focuses on performing specific tasks, whereas ‘General’ AI (AGI) can learn and solve problems in any environment; Narrow AI is ubiquitous (Google search, Amazon’s Alexa), but AGI is only hypothetical.

Narrow AI is trained to solve problems and adapt to change using the concept of machine learning;  algorithms analyse previous actions and autonomously improve when given new datasets making AI become more accurate.

In drug discovery AI can analyse vast quantities of data, and identify patterns too complex for humans to discern; it can then generate predictions, leading to the rapid and accurate identification of novel drug targets and lead molecules. AI uses natural language processing to bring together disparate information and datasets, providing insights that no single experiment could provide.

A key challenges in drug discovery is understanding the structure of the protein a drug could target; when discerned experimentally, the process is time consuming and expensive. Google’s DeepMind  recently launched AlphaFold, an AI platform that predicts protein structures with high accuracy, and a solution to a key bottlenecks in drug discovery.

Biological systems consist of highly complex networks of interactions, making it difficult to predict if a drug will have adverse effects. E-therapeutics (ETX.L) use AI to model and analyse these networks, and a representative simulation of a whole biological system helps translate therapies from laboratory to patient, reducing expensive clinical-stage failure.

‘AI is being touted to have a revolutionary effect on the drug discovery process, potentially mitigating attrition, accelerating development timelines or reducing cost. A plethora of partnerships have been struck between AI-vendors and drug developers, with a spate of fundraising for drug discovery platforms.’

Sean Conroy, healthcare analyst

The potential for faster and cheaper drug discovery has led to the founding of numerous start-ups; many have raised large amounts of investment and established partnerships with large biopharma companies.

BenevolentAI creates ‘knowledge graphs’ to connect related biomedical data from its large repository; they contain insights that humans would not be able to synthesise due to the complexity and volume of data. The information can identify drug targets, develop lead molecules and repurpose known drugs. BenevolentAI identified that baricitinib (an approved rheumatoid arthritis drug) could be used to treat COVID-19; FDA subsequently authorised its use to treat hospitalised patients. Collaboration combines BenevolentAI’s platform with AstraZeneca’s expertise and large datasets; in January, it announced the discovery of a novel target for chronic kidney disease.

Recursion (NASDAQ: RXRX) aims to make drug discovery faster and cheaper using machine vision to identify subtle changes in cell biology. This approach allows the company to rapidly analyse vast quantities of experimental data generated in-house by its automated robotic laboratory; it performs 1.5 m experiments each week. The company has four drug candidates in Phase I clinical trials and a partnership with Bayer to develop new therapies in fibrotic disease. Recursion completed its US$436m IPO on Nasdaq in April.

Overall, a new drug is estimated to cost over US$2.6bn and take at least 10 years ; although nascent, there is broad recognition of the potential of AI to improve the drug discovery process.

Since 2015 there have been around 100 new partnerships between AI services and the pharmaceutical industry; in November, Alphabet announced the launch of Isomorphic Labs, a spin-off of DeepMind, which aims to deliver an ‘AI-first approach’ to drug discovery.

“What is Synthetic Data?” >