UCSD researchers are featured in final teams of international cancer competition

Two researchers at UC San Diego are among members of a dozen teams selected as finalists in an international cancer competition that could net the winner up to $25 million.

Cancer Grand Challenges is a global research initiative made possible through a collaboration between Cancer Research UK and the U.S. National Cancer Institute. Thus far, the effort has contributed $400 million to 16 teams and engaged 1,200 investigators or contributors across the world.

From a record pool of 227 submissions, 12 teams were shortlisted as finalists for prizes this year. Winners will be announced in March at the Cancer Grand Challenges Summit in London.

Their projects seek to prevent, detect or treat cancer in six categories:

• Human and artificial intelligence collaborations

• Cancer avoidance

• The dark proteome

• Mechanisms driving mutational signatures

• The nervous system and cancer

• Rewiring cancer cells

Competing in the mutational signatures category is team CAUSE, led by Ludmil Alexandrov, a professor of cellular and molecular medicine and bioengineering at UCSD. Other collaborators on the team hail from the United States, Netherlands and United Kingdom.

This isn’t Alexandrov’s first shot at the competition. In 2017, when it was named the CRUK Grand Challenge, he was a collaborator for Mutographs, a team that analyzed “fingerprints” of environmental factors such as pollution or smoke that become etched on DNA of healthy and cancerous cells.

As one of that year’s winners, the team went on to conduct a global study on mutational signatures.

Now, Alexandrov and team CAUSE are focusing on understanding the basic molecular mechanisms that give rise to mutations, including those associated with aging, which Alexandrov says is the biggest risk factor for cancer.

“If we can understand these underlying mechanisms, the presumption there is that we can intervene and we can actually prevent cancer more effectively,” Alexandrov said.

Also vying for the top prize is Biologia Ex Machina, a team developing AI-driven “co-scientists” to aid human scientists in their cancer research.

Trey Ideker, a professor of medicine, bioengineering and computer science at UCSD, is a co-investigator for the team and is joined by peers from Spain, Switzerland, the U.S. and the U.K. Marinka Zitnik of Harvard Medical School is the team lead.

Following the public release in 2020 of GPT-3, a large language model from OpenAI, the team has “rigorously tested and controlled” LLMs, Ideker said.

The models’ ability to read all relevant literature and training shows significant promise for the technology, he added.

“Although there was, I think, a lot of appropriate concern about hallucination — that these things just kind of start making things up — for as many times as you worry, you also see inspired ideas out of these things and for what could be the underlying explanation of all your observations,” Ideker said.

“And you see it not just anecdotally. You see that time and time and time again these things are coming up with hypotheses that had a human thought it up, you’d think the human is brilliant.”

If Biologia Ex Machina gets the multimillion-dollar prize, it would do two things, Ideker said: hire a machine learning and software development team to use and push the technology ahead, and pursue experimental biology that the technology can help drive.

The ultimate question, he said, is whether computers will conduct all science a decade from now.

“The more I use these things and realize it’s doing a better job than I am at formulating deep hypotheses about cancer and underlying biological phenomena that drive cancer … I think it’s plausible that as a result of this Grand Challenges project, at the end of four or five years, we will have a system that’s basically posing and testing thousands of hypotheses every month,” Ideker said. “If you can do that, you’ve utterly changed the face of all of science.”

To learn more about the dozen finalist teams, visit cancergrandchallenges.org.