New Funding to Speed up Analysis of Patient Tumour DNA Data

Dec 11, 2017
Author: 
Jovana Drinjakovic

University of Toronto computational scientists have received new funding to develop faster and more accurate algorithms that can keep track of how tumours change over time in a bid to improve diagnosis and treatment.

Powered by artificial intelligence, the new software being developed by professors David Duvenaud (left) and Quaid Morris (right) could help doctors predict how a patient's cancer will respond to treatment.

 

Teams led by Professor Quaid Morris, of the Donnelly Centre for Cellular and Biomolecular Research, and David Duvenaud, of the Department of Computer Science, received a boost for their research that aims to drastically cut down—from weeks to seconds—the time it takes to build a tumour's “family tree” from its DNA sequence data in order to learn how the cancer has evolved and how it may respond to treatment.

Both Morris and Duvenaud are also members of the recently founded Vector Institute for artificial intelligence, launched with the support from U of T.

“We anticipate that tumour evolution data are going to be used clinically, but one of the main barriers to clinical use is computation time which can take days and sometimes weeks,” says Morris, who is also a professor in the Departments of Computer Science and Molecular Genetics. “But you can’t wait a couple of weeks for the software to run before making decisions about treatment, prognosis or even diagnosis.”

The $200,000 USD grant, which was awarded by the NVIDIA Foundation and the Silicon Valley Community Foundation, will go towards building a faster version of the artificial intelligence-powered software previously developed by Morris’ team. The algorithm works by scanning billions of DNA changes, or mutations, which have accumulated in tumours, to piece together an evolutionary history of a patient’s cancer.

As cancer begins to grow, its cells mutate much faster than the healthy ones, and in doing so acquire new qualities that allow them to spread and wreak havoc in the body. But not all mutations that are present in a tumour drive its growth—many are mere passengers, side-effects of the high mutation rate and therefore clinically irrelevant.

As mutations make their mark on cancer, researchers like Morris and Duvenaud can use tumour DNA sequence data to figure out which mutations came earlier and which ones came later in order to identify the ones responsible for the disease. This research could help tailor treatment and predict patient outcome as well as open new avenues for drug discovery.

Lots can also be learned from knowing the type of mutations and where in the tumour’s genome they accumulate in.

“Mutations accumulate differently in a breast tumour than in a liver tumour for example,” says Morris. “And types of mutations are different. Mutations caused by UV light in skin cancer are distinct from mutations caused by smoking that accumulate in lung cancer.”

These types of analyses could also reveal a cancer’s cell of origin, which is currently difficult to do for patients who learn they have cancer only after it has spread.

To speed up data analysis, Morris and Duvenaud plan to build a parallel computing platform

using multiple graphics-processing units (GPUs), a much faster type of computer hardware. Having first found major use in video games, GPUs have since become an essential component of deep learning, a form of artificial intelligence in which computers learn to recognize patterns from a large amount of data.

Nvidia Foundation is a philanthropic arm of NVIDIA, a technology company that makes some of the top GPUs on the market. In addition to Morris and Duvenaud’s teams, a team of researchers from the University of California in San Diego also received funding from the foundation, which has been supporting cancer research since 2011.

Follow us on Twitter to keep up with Donnelly Centre news.

Back to Top