Google’s parent company Alphabet has recently launched Isomorphic Labs, a venture utilising AI in its drug discovery practises.

The DeepMind AI system, which was developed by another subsidiary of Alphabet, will be used to power this discovery.

Announcing the plans in November of last year, Founder and CEO of both Isomorphic Labs and DeepMind, Demis Hassabis, published a blog post on the new company’s website. He wrote:

‘I’m thrilled to announce the creation of a new Alphabet company – Isomorphic Labs – a commercial venture with the mission to reimagine the entire drug discovery process from first principles with an AI-first approach and, ultimately, to model and understand some of the fundamental mechanisms of life.’

In the past, DeepMind has been used to rival the world’s champions in Go, an abstract strategy board game, with their AlphaGo machine. Their AlphaZero was also able to trump other powerful computers in chess, go, and shogi (Japanese chess) after playing itself for a few days, through AI reinforcement learning.

Unlike IBM Watson or DeepBlue, which were created for specific purposes, DeepMind is purportedly versatile and utilises its neural network to learn from experience, as opposed to being programmed.

Most recently, and perhaps most pertinent to Isomorphic Labs, was DeepMind’s AlphaFold2, which provided a solution for a 50-year long scientific challenge of protein folding. It enabled the prediction of the 3D structure of a protein through amino acid sequencing, down to atomic-level accuracy.

Likening natural systems to technology, Hassabis expounded:

‘At its most fundamental level, I think biology can be thought of as an information processing system, albeit an extraordinarily complex and dynamic one. Taking this perspective implies there may be a common underlying structure between biology and information science – an isomorphic mapping between the two – hence the name of the company’.

How does drug discovery currently work?

Drug discovery is currently an arduous, manual process. In the UK, it takes 10-15 years for a drug to get approved. Additionally, a mean of around £228,000 to £2 billion is required to discover and develop these new drugs. In the UK, only 1 or 2 of 10,000 compounds tested are cleared for commercial use.

In the UK, there are four main steps taken to introduce a new drug into the market.

Preclinical research occurs before testing, and biological knowledge of the illness being treated should be garnered, and extensive research is carried out. In the past, drugs were found by chance through the active ingredients in traditional medicines. However, with the current technology, scientists are able to pull from chemical libraries, and test compounds through reverse pharmacology – helped by the sequencing of human DNA. This compound is first tested on cells, then animals.

The first step, Phase I, tests the drug on a small group of healthy human volunteers, which will allow scientists to test for dosages and preliminary reactions.

Phase II consists of testing on subjects with the illness that can be treated with the drug. Lab studies may be done with placebos. This is done with a larger group of people, over the course of a few months, or even years.

If the prior stage is successful, Phase III involves testing on hundreds of participants worldwide, to ensure that results are seen in a diverse group of patients. This will take even longer, as it monitors how people react to the drug over time.

The last step is licensing, which is led by the regulatory body of the country. For England, Wales, and Scotland, this is the Medicines and Healthcare products Regulatory Agency (MHRA).

Although scientists have large libraries of genomic and chemical data to pull from when discovering and developing new drugs, the process of creating these combinations during the preclinical phase is time and resource-intensive. This is where AI comes in.

AI in drug discovery

AI is extremely adaptable and every year programmers are finding new ways to utilise it and optimise processes. Especially with neural networks such as DeepMind, the possibilities are endless – these digital ‘brains’ can interact with data from any field, and create effective strategies or fix any problems that arise. The computing power, unlimited information, and tirelessness are some of the main reasons why AI can be so effective. Menial, repetitive tasks can be done accurately and more swiftly than human workers. This technology is most useful in the preclinical stage.

In drug discovery, computers can scan libraries containing biological and chemical data, and suggest the most compatible compounds for testing – AI could completely automate idea formulation. Additionally, experiments can be digitally modelled for accuracy before being conducted in real life.

During experimentation, robots can be programmed to adhere to strict lab protocols, such as the stirring or shaking of chemicals a precise number of times.

Additionally, all manners of data can be captured without errors, such as temperature, humidity, quantities, and timing.

Each experiment can be analysed for efficacy – having a digital track record of experimentation makes the process much more manageable for researchers.

Isomorphic Labs is also planning on modelling more biological processes to better understand how these drugs affect patients.

Hassabis explained:

‘AI methods will increasingly be used not just for analysing data, but to also build powerful predictive and generative models of complex biological phenomena.’

Future promise

Accelerating drug discovery, Isomorphic Labs promises to aid in finding cures for some of mankind’s most devastating diseases. This marks a new era in man-machine modelling, new developments in AI provide a promise of better healthcare and further innovations in medicine. Although there is not much to be known yet about their future endeavours, the company will undoubtedly spur a paradigm shift within the industry.

As the Isomorphic Labs CEO concluded:

‘Biology is likely far too complex and messy to ever be encapsulated as a simple set of neat mathematical equations. But just as mathematics turned out to be the right description language for physics, biology may turn out to be the perfect type of regime for the application of AI.’

About the Author: Shadine Taufik

Shadine Taufik is a contributing Features writer with expertise in digital sociology and culture, philosophy of technology, and computational creativity.