Personalizing Cancer Treatments Through Data Science
Carnegie Mellon University is collaborating with the pharmaceutical company Novartis to develop new statistical methods that could improve personalized medical treatments.
By Stacy Kish
You are unique, and the best treatment for your illness may lie buried in a vast dataset of information gathered during a myriad of clinical trials. Mining that data to tailor the ideal therapy to treat your illness is one step closer to becoming a reality.
Joel Greenhouse, professor of statistics in the Dietrich College of Humanities and Social Sciences at CMU, is working with Novartis, a global healthcare company based in Switzerland, to develop and apply new statistical techniques to the company’s large drug outcome datasets with the goal of helping the company develop personalized treatments for diseases, including cancer.
“With the rise of data science, it became clear that pharmaceutical companies are sitting on a gold mine of data that have yet to be explored,” said Greenhouse. “Through this collaboration, we hope to bring together data scientists with subject-matter specialists to inspire not only new work in the statistical sciences but also help advance the pharmaceutical sciences in order to find treatments for people with serious diseases.”
Pharmaceutical companies are required to conduct clinical trials to demonstrate that a new drug is safe and effective before receiving approval by the U.S. Food and Drug Administration. Novartis sees great potential in mining these large datasets to answer questions beyond their original purposes.
“This collaboration will enable us to develop overarching predictive frameworks to integrate diverse models for disparate data types,” said Eric Gibson, global head of Biostatistics and Pharmacometrics, Novartis Pharmaceuticals. “This will give us actionable insights along the patient’s treatment journey that will increase the chances of treatment success. This could be potentially transformative in developing next generation personalized therapies.”
The CMU team will build models to identify patterns and characterize key causal links across imaging, genomic, clinical, biological and manufacturing datasets that have thus far remained elusive. One example of this approach explores the application of CAR-T, a personalized cancer treatment that modifies the patient’s own cells to attack certain cancers. The treatment developed by Novartis has been approved by the FDA to treat certain leukemias and lymphomas.
The agreement enables graduate students and faculty to develop and apply novel statistical approaches for combining information across multi-dimensional drug datasets. Robin Dunn, a doctoral student in Greenhouse’s group, is analyzing data on the Novartis CAR-T studies. In this thread of the collaboration, she is constructing a multi-state model to evaluate therapy success based on data collected at various stages of diagnosis and treatment, including the initial blood draw, the manufacturing process, the CAR-T infusion, early response to the therapy and longer-term health outcomes.
“Our modeling aims to incorporate patient-to-patient variability in the study,” said Dunn. “This work will provide insights into critical attributes of the initial blood sample, important factors in the manufacturing process and characteristics of the best candidates for CAR-T therapy.”
Novartis plans to implement the results of this collaboration to improve drug development that could transform how medicines and treatments are discovered, developed and commercialized.
“CMU is famous for working on problems that matter,” said Greenhouse. “I see this collaboration as another opportunity for the university to make a contribution to the larger world by helping bring the best therapies to patients.”