Skip to content

Drug combinations that help breast cancer patients discovered using data science

drug-interactionsWhen computer scientist Andrew A. Radin came to Stanford University School of Medicine and enrolled in a biomedical informatics course, he was just there to learn.

For a class project, he made a software platform that used a range of different methodologies to find diseases that might be helped by already existing drugs. Each methodology had deficiencies, but by aggregating the predicted associations between drugs and disease, Radin says, he could accurately predict which drugs might be helpful for which diseases.

“The thing worked awesome!” he said in a talk he later gave at Stanford.

Radin got a good grade on the project and was planning to publish the work as a scientific paper. But a venture capitalist he knew told him to forget the paper and file a patent and start a company instead. Before long, it was clear that the all-consuming startup, twoXAR, was keeping him from finishing his course work at Stanford.

Biomedical data scientist and physician Nigam Shah, MBBS, PhD, who was teaching the course, took Radin aside and asked what was wrong. Radin admitted what he was up to, and the two quickly agreed the new company was more important. But because Radin wasn’t doing his work, Shah didn’t hesitate to give him an F.

There are no hard feelings, though, and today both agree it was “the best F ever!”

Friday, the pair came full circle with a team effort that revealed combinations of drugs that could reduce mortality in breast cancer patients. The research appears in the Journal of the American Medical Informatics Association.

As a Stanford Medicine press release explains.

'We looked at all the noncancer drugs that breast cancer patients were on,' said Shah. 'People have other things going on in life. They might have hypertension, they might have high cholesterol or diabetes. They would be taking drugs for those as well. So the question we were asking was, do any of the drugs they are taking associate with better outcomes for breast cancer?'

The team looked at data from nearly 10,000 adult women diagnosed with breast cancer between 2000 and 2013, of whom about 12 percent died within five years of the diagnosis. The team examined 294 drugs in more than 43,000 pairwise combinations. Specifically, they looked for combinations of drugs in which the beneficial effect on survival was greater than the effect of either drug by itself.

'So we ran the analysis, and we found a few drug combinations that seemed to associate with better survival,' said Shah.

Then Shah and his team turned to Radin for confirmation using his quite different approach and independent datasets. “Lo and behold, we both came up with the same answer,” said Radin. The two approaches are different ways of finding new drugs for old diseases.

And Radin finally got his name on that academic paper.

Previously: Digital clues for healthier drug combinations: Stanford's Russ Altman at TEDMED and Patient data shows that common prostate cancer treatment likely doubles risk of dementia
Photo by Shutterstock

Popular posts