How ‘-omics’ data can revamp treatment of diseases

Molecular data can help inform clinicians’ care decisions and give advance warning of illness, says Anthony Philippakis, MD.

Healthcare professionals can draw lifesaving conclusions from -omics data sets—molecular data from genomics, proteomics and metabolomics—to help inform their clinical decision-making and improve patient care.

What distinguishes -omics data from previous molecular biology research is the ability to express all of a cell’s proteins, rather than a single protein, says Mark DePristo, head of deep learning for genetics and genomics at Google. “You collect a lot of data very quickly because you're not selecting for specific proteins you want to cite,” DePristo notes.

DePristo and Anthony Philippakis, MD, the chief data officer of the Broad Institute, a Harvard and MIT biomedical and genomic research center, will be among the experts discussing this topic at the HLTH session “Discovering Meaning From Massive -Omics Data Sets” scheduled to begin at 3:20 p.m. on Tuesday, May 8, at the Aria in Las Vegas.

Philippakis, the moderator of the panel, says the session will cover four areas:

Gaining insights from genomic and clinical data before disease strikes
Genomic and clinical data could provide insight to identify diseases before patients gets sick. In fact, metabolites in bodily fluids bring insights on the risk of various diseases, and these insights could make preemptive care possible before disease afflicts a patient. The -omics data might provide insight on the chances that a patient may suffer a heart attack before they are diagnosed with cancer. The challenge is to be able to pull out that data from the data sets carrying those signals.

“We have millions and millions of data points per person, potentially millions of people soon—how do you go and find spots in the genome that confirm your risk for heart disease?” DePristo asks.

When a patient has lung disease, doctors are able to prescribe a chest X-ray to scan for cancer, spot a nodule and then operate on it before the nodule metastasizes. “However, the reality is that when you try to do this systematically on a population level, it's a lot harder,” Philippakis explains. “So why do we think that it will be any easier with new data types than it was with imaging data types in the past? I think that's a very important question for the field.”

Examining the organizational culture of medicine and tech
The technology and medical communities are both “incredibly innovative, incredibly important to the future of humanity, but their cultures and patterns for what success looks like are wildly different,” Philippakis says. The medical industry values experience and is more “hierarchical,” while the tech industry has a “flatter” organizational structure and is more open to experimentation than healthcare, he explains.

“And yet, we’re at a time and place where these two worlds need to come together in a synergistic fashion,” Philippakis says. “How do you build the right cultures and organizational structures that bring the best of both?”

Developing a learning medical system
Today, the data from clinical care and medical research live separate from each other. The healthcare industry needs a learning medical system that collects data on patients during the process of providing care and then uses this data to help care for other patients.

“That data is used to fuel discoveries and then is fed back into the [population health] system to inform the care of the next generation of patients,” Philippakis explains. Machine learning is an example of a learning medical system, and the Broad Institute has developed a software platform called the Genome Analysis Toolkit to gain insights from genomic and clinical data.

The role of artificial intelligence in medicine
AI holds tremendous potential in the healthcare field as a way to match patients to chemotherapy, Philippakis notes. “As a cardiologist, I certainly see the potential for AI in the care of cardiovascular patients,” he says.

Machine learning tools can perform this type of binary decision-making for predictive analytics, like whether or not to put in a valve, start a blood thinner or put in a stent. “Right now, we do these by intuition and guidelines, but we don't have sophisticated machine learning tools,” Philippakis adds.

The HLTH session will also cover whether AI could play a role in radiology and pathology. “I think we'd like to have a little bit of a discussion on what are the areas of medicine that will be most transformed by AI, and is the state of technology really what’s limiting us, or is it more the business models and incentive structures of the healthcare system?” Philippakis says.

More for you

Loading data for hdm_tax_topic #better-outcomes...