AI Finds Anti-Cancer Drug Candidates With Quantum Computing

Quantum-classical AI finds molecules targeting “undruggable” cancer proteins.

Posted February 10, 2025 | Reviewed by Davia Sills

Innovative technologies such as artificial intelligence (AI), machine learning, and quantum computing can accelerate the design and discovery of novel drugs to fight cancer. A new peer-reviewed study published in Nature Biotechnology, led by the University of Toronto and Insilico Medicine, demonstrates how AI can create anti-cancer molecules for proteins that previously could not be targeted pharmacologically by leveraging quantum computing.

“In my opinion, quantum computing is in the same state today as generative AI in chemistry was in 2015 to 2016, when my team decided to go all-in on it,” said Alex Zhavoronkov, founder and CEO of Insilico Medicine. “We know it is early; we know that we will not get the benefits right away, but we want to push the boundaries for ourselves and for the entire industry, specifically focusing on what we know best—chemistry and biology.”

The Global Problem of Cancer

Cancer is a leading cause of death worldwide, second only to cardiovascular diseases, according to Our World in Data . Globally, by 2050, an estimated 35 million cancer cases are projected, according to the Global Cancer Facts & Figures 5 th Edition by the American Cancer Society (ACS), and roughly 1 in 5 Americans will develop cancer in their lifetime.

Cancer is a broad spectrum of diseases in which healthy cells become cancerous due to uncontrolled cell growth that can spread. According to the National Cancer Institute (NCI), cancer is a genetic disease caused by mutations in the genes that manage our cells. A genetic mutation is a change in the DNA sequence. An oncogene is a mutated gene with the potential to cause cancer.

It is important to note that even though cancer is a genetic disease, that does not necessarily equate to it being a familial or hereditary disease. In fact, according to the NCI, a majority (90 percent to 95 percent) of cancers are caused by mutations and are non-hereditary (also known as spontaneous), where lifestyle and environment impact cancer risk. For example, cigarette smoking , tobacco use, certain infections, radiation, and immunosuppressive drugs used for post-organ transplant are factors known to increase cancer risk, per the National Cancer Institute.

What Is KRAS and Why Target It?

For this new study, the team of researchers led by the University of Toronto and Insilico Medicine created a hybrid quantum-classical generative AI algorithm to create small molecules to target the KRAS (Kirsten ras oncogene homolog) gene. KRAS genes belong to the class of genes called oncogenes. When an oncogene is mutated, it has the potential to cause cancer. In the early 1980s, scientists identified the first human oncogenes. In 1982, Harvard Medical School researcher Channing J. Der and other scientists at the lab of Professor Geoffrey Cooper published their discovery of the role of Harvey and Kirsten sarcoma viruses in human bladder and lung carcinomas.

KRAS gene mutations account for an estimated one-third of all cancers, according to the National Cancer Institute, and can be found in 90 percent of pancreatic cancers, according to a 2024 precision oncology study published in Surgical Oncology Clinics of North America by medical doctors Newhook, Tsai, and Meric-Bernstam. KRAS gene mutations are also linked to 40 percent of colorectal cancers and 32 percent of lung cancers, according to Dr. Shubham Pant, a professor at The University of Texas MD Anderson Cancer Center.

Why Quantum Computing?

The emerging field of quantum computing has the potential to be exponentially faster than today’s classical computing (binary computing) that uses binary logic and bits (zeros and ones) to process information.

“When in 2017-2018 we showed our first Entangled Conditional Autoencoder (ECAAE) and a similar experiment to this quantum work showing that we can synthesize molecules out of generative AI that work experimentally, nobody listened. Then, in 2019 we published our Generative Tensorial Reinforcement Learning (GENTRL) system and synthesized and tested the first molecules and tested in mice, it set the stage for the entire field and many other people started doing the same. This quantum work is similar to ECAAE in 2018.”

Quantum computers apply quantum mechanics to solve complex problems and use quantum bits (qubits) to store and process data. Not only can qubits have a state of zero or one, but also a weighted combination of zero and one simultaneously.

“In 2026-27, many hyperscalers, including Microsoft and Amazon, will have their quantum machines scale and open for quantum as a service. In China, you can also already buy time on a real quantum machine as a commercial service. We will closely monitor these groups to ensure our software is tailored for drug discovery applications on quantum.”

Hybrid Quantum–Classical AI

The researchers created a quantum-assisted AI algorithm with a classical computing LSTM algorithm with a quantum generative AI model. The team generated a training data set of over 1.1 million molecules using 250,000 molecules screened from 100 million molecules using VirtualFlow and 650 experimentally validated KRAS inhibitors. The researchers used their hybrid AI model to generate new candidate molecules for targeting KRAS and Insilico Medicine’s generative AI engine Chemistry42 to predict the top 15 ones to assess in a lab, of which two, in particular, have enormous potential as future KRAS inhibitors. The researchers found that their hybrid quantum-classical model outperforms classical models, and their findings are an important proof-of-concept.

“Even though our hybrid quantum-classical model outperformed conventional machine learning models in generating structurally diverse and synthesizable drug-like compounds, with a 21.5 percent higher success rate in meeting drug design criteria, in this paper, we are not claiming that we are faster, cheaper, or better than GPU, we are just showing that it is possible,” said Zhavoronkov.

Share this post Facebook Bluesky Linkedin Email

There was a problem adding your email address. Please try again.

By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy

Cami Rosso writes about science, technology, innovation, and leadership.

Get the help you need from a therapist near you–a FREE service from Psychology Today.

This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.