HPC allows UI cancer researchers to simulate tumor development

Dr. Donghai Dai, associate professor of obstetrics and gynecology

Uterine cancer is the fourth most common cancer among women, and the most common gynecological cancer. Researchers at the University of Iowa are using high-performance computing (HPC) to investigate how a tumor develops in normal uterine epithelium.


The majority of cancer research presumes that every cell in a tumor is driven by the same genetic alterations and follows same pathway to malignancy. Dr. Donghai Dai, associate professor of obstetrics and gynecology in the Carver College of Medicine, doesn’t think that’s the case. He believes the billions of cells that make up a tumor may each have their unique mutations that cause them to deviate from normal cell behavior.


So, Dai and his team are taking a single-cell approach to studying the development of tumors. With complex mathematical models and the Helium computing cluster, administered by Information Technology Services, they can run simulations on millions of representative uterine cells each day. The researchers determine the fate of the cells through incorporation of the combined effect of numerous random mutations and varying hormonal stimulations.


“The prevailing approach to study the mechanism of oncogenesis is by examining a tumor tissue after it is surgically removed from a patient, and retrospectively determining how and why the cancer formed,” Dai said. “However, it’s difficult to determine exactly what caused the cancer because usually there are multiple mutations and other compounding factors that may have collectively contributed to it.”


With computational modeling, Dai and his team study tumor development prospectively. They begin with normal cells, manipulate variables such as hormone levels and mutational effects for each individual cell, and observe the probability of tumor development. This allows them to pinpoint exactly what changed in those cells.


“There’s no biological way to dissect, among hundreds of mutations, which one contributed to the formation of a tumor,” Dai said. “But we can do that with the computer simulations.”


Because the formation of a tumor is a low-probability event, the researchers must run massive numbers of simulations. Without HPC, a simulation of uterine cells across the lifetime of one woman takes about 15 days. Now, they can run as many as 20 of those simulations per day.


“We need the power of the HPC cluster to run these simulations in a timely manner, and to run enough of them that we are statistically confident in the results,” said Cory Howk, a postdoctoral researcher involved with the study.


In just nine months of HPC-assisted research, the team has made several discoveries.


For example, they believe menstruation provides a natural defense against tumor development in the uterus. Post-menopausal women are 10 times more likely than young women to develop uterine cancer, but the UI team found that young women actually have more cancerous cells in the endometrial tissue. Dai said it’s likely that young women shed live cancerous cells along with the surrounding dead tissue; when women stop having a period, they lose this defense mechanism.


With computational modeling, the UI scientists have been able to examine a cell's behavior along a spectrum, rather than simply classifying cells as cancerous or noncancerous.


“The biological reality is that it’s not that clear cut, so we treat it as a continuum of behaviors,” Howk said. “We’ve learned that the strength of the mutation is important in determining whether a tumor will develop. Another key factor is the cell’s location on its developmental pathway from a proliferative cell immediately born from a stem cell to a terminally differentiated and senescent cell. The earlier a mutation occurs, the higher the likelihood that its descendants will be cancerous.”


This team is reporting their findings that the high prevalence of certain hereditary cancers in affected families, such as Lynch Syndrome, Cowden Syndrome and Li-Fraumeni Syndrome, is primarily due to the occurrence of some mutations in early stage of tissue development (germline mutations) and their presence in every cell in our body.


The next step is to look for common ancestors within tumors – whether the cancerous cells descend from the same initial progenitor cell, or if there are more recent ancestors. If they descend from the same parent cell, researchers will investigate the properties of the cell that spawned the tumor.


Dai said the research based on computer simulation of tumor development in human tissues has great potential to capture the entire process of tumor development and will be a valuable addition to a cancer research field currently still dominated by experimental approaches.


“The supercomputer resources give us the ability to do work that we could not accomplish with bench research alone,” Dai said. “These resources give us an increasingly powerful tool to conduct a dream experiment in cancer research by following the natural course of tumor development in living persons through simulation.” 


Illustration:  University of Iowa cancer researchers use this illustration to show the tumor development process they are simulating with the help of high-performance computing.