As data science changes the way drugs are discovered and developed, pharmacy schools are exploring the possible benefits for research and patient care.
By Joseph A. Cantlupe
At the Quantitative Biosciences Institute at the University of California San Francisco School of Pharmacy, there were hints in January that the world was changing. As the coronavirus pandemic began to emerge on the West Coast, much of academia was on vacation. Inside the QBI scientists scrambled, trying to figure out something massive and unknown: what made the coronavirus tick. Over the next few months, the lab went full bore to examine the intricacies of the coronavirus, spilling its research across three countries and enlisting more than 200 scientists.
QBI Director Dr. Nevan Krogan formed the QBI Coronavirus Research Group, eyeing multiple research projects related to COVID-19. Researchers pooled their expertise in biochemistry, virology, and structural, computational, chemical and systems biology to understand the intricacies of how the coronavirus effectively undermines human cells to replicate itself rapidly, enabling spread to others, and what could be done to thwart it. They began to explore immense possibilities in their research: rapid diagnosis using gene-editing technology to track the spread and evolution of COVID-19, and diagnosing infected patients with no or minimal symptoms.
The researchers tapped into the world of data science, which is constantly evolving and changing the way drugs are discovered and developed and how treatments are delivered to patients. Pharmacy schools are moving forward to be part of this trend, especially in research, where data and computational methods are becoming a major part of pharmaceutical and health sciences.
For Krogan, his around-the-clock work with his research team was an opportunity to embrace research unburdened by bureaucracy and focused on a team approach welcoming to students. “We had a foundation for collaboration, which doesn’t happen overnight. It was expedited in an exciting way,” Krogan said. Noting the progress that the multidisciplinary team made, he said similar academic work “would have taken a year, but it came together in a few weeks.”
As the QBI Coronavirus Research Group set out to uncover the human proteins enabling the coronavirus to spread, it looked at how human and virus proteins interact and studied the clusters they formed. In the meantime, the team identified at least 75 over-the-counter prescription and development-stage drug compounds that they said had the potential to target cellular proteins that are possibly “hijacked” by the virus to promote its spread. The Krogan team, which included 38 scientists, made maps of cells both in healthy and disease states and disseminated the information around the world for other researchers to examine.
Data Opening Doors
“It’s all data-driven, what we piece together in this pipeline, and show how fast it can move,” Krogan said. “We interface and try to find the right targets to have an integrative suite of tools for the underlying biology. It’s sharing across the board and it is done in a more fully open way across academia.“
At the University of Florida College of Pharmacy, Dr. Steven M. Smith, assistant professor, focuses on heart disease, stroke prevention and hypertension control. He also is beginning to examine the coronavirus and possible links to hypertension and its treatment. In his research, Smith aims to “improve the way we inform treatment decisions and interact in multidisciplinary teams in the future through data visualization and clinical support.”
Smith, who is in the pharmacotherapy & translational research department and was AACP’s second NAM Fellow, also runs a family medicine fellowship at Florida and is involved in training. Data flows through many areas of healthcare. It involves interactions with patients, providers and insurers and includes medical records, administrative claims, such as billing and patient pharmacy data related to medication adherence, and even patient-reported outcomes. Fundamental issues in his work include data collection and curation, standardization, privacy and interpretation. Emerging technologies, like artificial intelligence and advanced data visualization, have to account for these issues, he said.
“If you ask 10 different people what data science is you’ll get 20 different answers,” said Smith, noting a longstanding debate over the issue. “People who call themselves data scientists don’t necessarily agree on what that means. I think pharmacy schools are recognizing that it’s important to start integrating this into the profession. Some have had this focus for a while, but others are just beginning to appreciate the importance of bringing data science into pharmacy.”
At the University of Florida, data science and related concepts are not a major focus of the professional Pharm.D. program but are taught in graduate Ph.D. and master’s programs. The graduate programs require students to get coursework in fundamentals of inferential statistics, data analysis and interpretation, all related to but not explicitly framed as data science, Smith said. The university is advancing a large-scale initiative focused on what officials say is making the school a leader in artificial intelligence research and training, including the use of supercomputer technology.
Smith is also interested in using big data sources for developing prediction models, especially related to hypertension studies. The results of blood pressure measurement may differ in so many ways, depending on where the clinicians are: in a doctor’s office, a home or a hospital. Making sense of these data can be challenging. Smith works closely with the Patient-Centered Outcome Research Institute’s PCORNet that aggregates real-world patient data, linking electronic health records, health plan data and patient-reported outcomes within a standardized common data model. PCORnet serves as a repository for data on nearly 70 million Americans receiving care across the U.S. and provides infrastructure for large-scale pragmatic clinical trials, he said. “We have started to push the boundary of the scale, speed and kinds of data that can be collected and use that to try to rapidly answer questions that weren’t previously possible, such as whether certain treatments are better or worse for patient groups that are often excluded from phase 3 clinical trials.”
Building on Existing Capability
Dr. Allen Flynn, assistant professor in the Department of Learning Health Sciences, and a research analyst and technology lead at the University of Michigan Medical School, discusses data science as “essentially the automation of algorithmic data modeling.” And that, he said, is steeped in statistics. At a place like the University of Michigan, “data science is getting a lot of attention and support campuswide these days.”
Pharmacy schools can build on their data science capability, beginning with statistical training, by adding a course in data science and machine learning, Flynn said. “One of the challenges, which is also the case in statistics courses these days, is that data science is pursued with sophisticated software tools that students also need to learn,” he pointed out. As in many areas of healthcare, data can be scattered in different uses, complicated and difficult to scrutinize. “In healthcare, because data are highly fragmented and often productized and commercialized, it can be difficult to find sufficient data to get significant value from data science studies.”
Still, substantial strides are being made, especially with large image data sets, big electronic health record data “and with enormous quantities of streaming data from physiologic monitors,” he said. Michigan Medicine is working to implement models from data science to help predict “sepsis, readmission, [electronic records] utilization and the downward trajectory of clinical worsening.”