Skip to Content
Artificial intelligence

An AI-Driven Genomics Company Is Turning to Drugs

Deep Genomics aims to develop drugs by using deep learning to find patterns in genomic and medical data.
Niv Bavarsky

The next blockbuster drug could be developed with help from machine-learning techniques that are rapidly spreading from AI research to pharmacology labs.

Deep Genomics, a Canadian company that uses machine learning to trace potential genetic causes for disease, announced Tuesday that it’s getting into drug development. It joins a growing list of AI companies betting that their techniques can help produce powerful new drugs by finding subtle signals in huge quantities of genomic data.

Deep Genomics was founded by Brendan Frey, a professor at the University of Toronto who specializes in both machine learning and genomic medicine. His company uses deep learning, or very large neural networks, to analyze genomic data. Identifying one or more genes responsible for a disease can help researchers develop a drug that addresses the behavior of the faulty genes.

Until now the company has focused on scouring the genome for hard-to-detect mutations that might have a causal relationship with a particular disease. The company will focus, at first, on early-stage development of drugs for Mendelian disorders, inherited diseases that result from a single genetic mutation. These diseases are estimated to affect 350 million people worldwide.

The rush to apply AI techniques to medicine and drug development is partly driven by the emergence of powerful new algorithms, but also by cost-effective new ways of sequencing whole genomes, the entire readout of a person’s DNA. “There’s an opening of a new era of data-rich, information-based medicine,” Frey says. “There’s a lot of different kinds of data you can obtain. And the best technology we have for dealing with large amounts of data is machine learning and artificial intelligence.”

Deep learning has emerged in recent years as a very powerful way to find abstract patterns using large amounts of training data. It has proved especially valuable for speech recognition and for classification (see “10 Breakthrough Technologies 2013: Deep Learning”). The approach is now rapidly finding new uses in fields including medicine, where it offers a way to spot signs of disease in medical images and has shown potential for predicting disease from patient records.

Frey, who trained as a computer scientist and studied at the University of Toronto under Geoffrey Hinton, a key figure in the development of deep learning, says Deep Genomics will seek to partner with a pharma company on drug development. But he adds that the company offers key expertise.

“There’s going to be this really massive shake-up of pharmaceuticals,” Frey says. “In five years or so, the pharmaceutical companies that are going to be successful are going to have a culture of using these AI tools.”

Deep Genomics has also specialized in finding elusive and less direct disease triggers in a person’s genome. The company has, for example, published work showing how deep learning can help identify patterns in DNA that might contribute to diseases such as spinal muscular atrophy and nonpolyposis colorectal cancer.

Stephan Sanders, an assistant professor at UCSF School of Medicine in San Francisco who specializes in using genomics and bioinformatics to study disease, says deep learning could help with drug development by finding patterns in sparse pathology data combined with large genomic data sets. “We have vast amounts of data; three billion data points per individual,” Sanders says. “What we have less of is the other end: clean data of phenotypes or outcomes.”

Several other companies are seeking to apply machine learning to drug development. These include BenevolentAI, a British AI company, and Calico, a subsidiary of Alphabet.

Ken Mulvany, founder of BenevolentAI, says his company is focused on diseases of inflammation and neurodegeneration, orphan diseases, and rare cancers. And it aims to tap into largely unused research data. “Developing medicines is still a very lengthy, risky, and expensive process with high rates of attrition,” Mulvany says. “[But] there is an enormous amount of untapped data located in pharma R&D organizations without any plans to develop it.”

Deep Dive

Artificial intelligence

This new data poisoning tool lets artists fight back against generative AI

The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models. 

Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist

An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.

Driving companywide efficiencies with AI

Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.

Unpacking the hype around OpenAI’s rumored new Q* model

If OpenAI's new model can solve grade-school math, it could pave the way for more powerful systems.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.