November 10, 2015
Over the past decade tumor-modeling group at the Institute for Computational Engineering and Sciences (ICES) has been using computational methods to model cancer treatment.
Led by ICES Director and ASE/EM Professor Tinsley Oden, the group’s focus on modeling prostate cancer and how it responds to laser ablation has helped refine laser therapies to more effectively destroy cancer cells while minimizing damage to healthy surrounding tissue.
But the team’s latest research venture into cancer has a different objective—instead of modeling how cancer cells respond to therapy, they’re first seeking to model how they thrive in their native environment.
By capturing the complex physiology of cancer in a computational model, the team is working to develop tools that they and other researchers can later use to model patient-specific cases, and test various kinds of treatment.
“If you have calibrated a mathematical model you can attempt to design something for specific patients and that is the way much of modern medicine is now moving,” Oden said.
Along with Oden, two researchers in particular are helping to drive the group in this new direction, Thomas Yankeelov and Nichole Rylander.
Yankeelov, the director of cancer imaging research at Vanderbilt University’s Institute for Imaging Science and the Vanderbilt-Ingram Cancer Center, is working to integrate advanced cancer imaging methods with computational modeling of tumor growth and treatment responses. Rylander, an associate professor in UT’s Department of Mechanical Engineering and ICES faculty member, is studying cancer in the lab to both inform the computational models and test their results through experiments.
The Rise of HPC and DNA
Historically, cancer research has been the domain of scientists who study the disease through cells in a dish, animal models or patients.
But in the past decade, computational scientists and their models have become an increasingly important player in the process. The rise is due to increasing abilities of genetic sequencing and high performance computing, Oden said.
While DNA sequencing has shed new light on cancer—a disease that’s rooted in damage to the genetic code—high performance computing is enabling researchers to integrate this knowledge into multiscale models. These models are built by combining the complex network of physical and biological interactions that occur in cancer from the genetic, cellular, and tissue levels.
Because these models are built from the molecular level up, initial errors compound if they’re allowed to propagate through the model. Therefore, to know how trustworthy a model is, it’s essential to quantify the uncertainty present in results at each step of the way, Oden said.
Oden and the tumor-modeling group have spent years developing and refining uncertainty qualification algorithms, as well as algorithms for physiology and growth, for models of prostate cancer growth and treatment. In his latest venture he plans to integrate ICES codes with Yankeelov’s imaging and modeling methods.
ICES Tumor Modeling Group
From Image to Model
Yankeelov’s educational background includes degrees in applied mathematics, physics, and biomedical engineering. He applies his knowledge from across these fields on a singular goal: studying cancer.
“We study cancer exclusively. I’m told there are other diseases but I’m not certain what they are,” Yankeelov joked.
He primarily studies breast cancer.
His work includes improving imaging techniques that non-invasively capture details about a tumor’s form and composition. He then uses these images to create computational models of a tumor that can be used to predict its physiology and future behavior.
Yankeelov has developed a “family” of computational models of varied complexity that can be applied to imaging data. It’s difficult to know which model is best suited for handling a particular tumor image, Yankeelov said, in large part because of uncertainties present in the imaging and modeling data. By collaborating with Oden he’s hoping to improve the selection feature.
“One of the very attractive things about Dr. Oden’s group is they can take an agnostic point of view on what model is the best because they developed these methods for model selection that given a family of models can determine which one is the best given the data,” Yankeelov said. “And that’s something that’s not easy to do.”
A cancer model’s behavior is driven by software made up of equations and algorithms. To ensure the relationships and data contained in the code represent reality, the tumor-modeling group is turning to Rylander and her tumor-engineering laboratory.
A Platform for Cancer
To understand cancer behavior in its natural environment Rylander grows tumors in lab environments that emulate important features of a living body.
These tumor “platforms” include a cell matrix that promotes naturalistic tumor development and includes microfluidic channels that mimic blood vessels and assist in the delivery of important molecules and products, such as oxygen and proteins, to the cells.
“We use tissue engineering to create tumors and vessel structures that mimic the microenvironment of tumors,” Rylander, who mainly studies breast cancer, said.
Rylander has worked with Oden on cancer models before. She earned her Ph.D. in biomedical engineering developing similar platforms for the prostate tumor research. Now, after eight years on the faculty of Virginia Tech, Rylander has returned to UT and is again a part of the ICES tumor-modeling team.
The tumors and microenvironments she and her lab creates are closely monitored to provide data to improve computational models. Different cell types are tracked by tagging them with florescent dyes, genetic products are examined as they’re made using real time polymerase chain reaction, and forces such as the shear stress and velocity of fluid moving through vessels is measured using novel Imaging methods.
As a Ph.D. student her experiments helped illuminate how heat shock proteins released by prostate tumor cells after laser exposure helped repair cellular damage leading to tumor survival and recurrence—a feature that the computational model was not accounting for.
Oden says its insight like this that will help improve the function of future models.
“At the institute we developed a significant collection of computer programs, computer algorithms, and libraries that are coded and working to depict a wide range of phenomena associated with the growth and death of living tissue,” Oden said. “What’s been missing is well orchestrated in vitro experiments to calibrate these models.”
Future Plans
Oden, Rylander, and Yankeelov are planning on combining their expertise to develop computational models and experimental platforms for liver cancer. They recently submitted a funding proposal to fund these studies to the National Institutes of Health.
Their work, for the moment, will take place purely in the research sphere. But the goal is for computational models—for cancer, as well as many other diseases—will become a clinical tool.
Between the tumor-modeling group’s new ventures, and its many centers devoted to biomedical research, ICES is preparing itself to be a leader in developing the technologies that could one day make computational medicine a literal life saver.
“We’ve been promoting this work for over a decade,” Oden said. “Now you can see it’s catching on.”
