Director, Head and Neck/Oral Oncology Program
U of Connecticut Health Center's Neag Comprehensive Cancer Center
Douglas Peterson directs the Head and Neck/Oral Oncology Program at the University of Connecticut Health Center’s Neag Comprehensive Cancer Center. Atthe recent American Society of Clinical Oncology meeting, he presented results from a computer simulation of a complication of high-dose cancer therapy called mucositis, which is injury to mucosal membranes.
The simulation used the Virtual Cell platform, a distributed software application and modeling environment that the National Resource for Cell Analysis and Modeling is creating for quantitative cell biology research. The resource is housed at the UConn Health Center’s Center for Cell Analysis and Modeling, or CCAM, directed by Leslie Loew.
Virtual Cell offers a java-based interface tool that allows users to build complex models using their experimental parameters. The software converts biological descriptions into differential equations, while numerical solvers generate software code to enable simulations, which are downloadable in various formats. The software is available here.
At ASCO, Peterson and his colleagues showed how Virtual Cell simulated the dynamics of an inflammatory mucositis-reaction network. The simulation drew on a set of human COX-2 pathway data curated from the literature as well as data from oral mucosal biopsies taken from pre- and post-bone-marrow-transplant patients. It also incorporated feedback loops and accounted for the response of a number of other inflammatory species and their interactions, all of which were captured to create the mathematical model.
Cancer therapies like chemotherapy or radiotherapy may defeat tumors, but oral mucositis is a frequent and painful complication of treatment for many patients. What starts out as soreness can escalate to more intense inflammation and ulcers, which can make it impossible to eat or even swallow pills. The condition also increases infection risk, which could become fatal for cancer patients whose immune system has been weakened by chemotherapy or radiation.
The pathway to be modeled is complex: Mucosal injury starts with skin cell death and is likely exacerbated when the free radicals generated by chemotherapy-induced cell damage join the mix. The activated immune system calls pro-inflammatory cytokines to the scene, further injuring the mucosa and activating pathways that amplify the injury.
When Peterson’s team ran the simulation and compared the results with the compiled experimental and literature-derived data, they found that Virtual Cell predicted eight of nine dynamic inflammatory responses: COX-1, COX-2, TNF-alpha, PGE-2, IL-1beta, IL-6, PGI-2, TXA-2, and NF-kappaB.
Peterson originally followed in his grandfather’s footsteps to dental school, intent on not becoming involved in a huge medical enterprise. His plans shifted along the way when a pathology class captured his interest. After completing dental school he embarked on a five-year effort to study tumor immunobiology, which included a fellowship at the National Cancer Institute.
He was steering toward a career in basic research when he said he realized he had more of a translational research bent, enjoyed working with patients, and saw there was a gap in the professional management of mouth-related complications of cancer therapy.
BioInform spoke to Peterson upon his return from ASCO. The following is an edited transcript of that conversation.
What has motivated you to want to model mucositis?
I started out more clinically and translationally; I am doing a lot of homework to catch up on the computational side. We have a great team of engineers, protein chemists, biological scientists, translationalists, and clinicians. We work together well as a team conceptually — I am very interested in the bridge effect from the lab to the clinic but also across the disciplines.
Some of these [treatment-related] toxicities are so complex in their cause, in their diagnosis and treatment [and research] that it requires a multi-professional team.
Somewhere between 5 [percent] and 40 percent of all patients receiving chemotherapy in standard dosages, and 75 [percent] to 80 percent of patients on either high-dose chemotherapy regimens prior to bone marrow transplantation or radiation for head-and-neck cancer, are afflicted by mucositis.
What information flows into the mucositis network you are developing?
Over the years we have developed a computational model based on a number of sources, including literature outside of mucositis, such as literature on COX-1 and COX-2. There have been other individuals who have contributed to molecular models; a lot of it is theory. What we are trying to do is to move from the theoretical framework into research and generation of new knowledge.
The mucositis network derives from the literature on inflammation and the literature on mucosal injury in cancer patients. The network itself is the model of our computational biology … developed through [a series of] kinetic equations. ... So it is a mathematical approach to equation modeling that generates this network.
Once you have developed the model, Virtual Cell allows you to run a series of simulations based on rate constants. Our goal is on a number of levels: One is to take the current evidence and integrate the data from the literature and from our laboratory so you get a model based on actual data in mucositis.
In doing so we have learned there are significant gaps in the literature. There just isn’t evidence in many of these areas of mucositis. … Already we have generated some new hypotheses and the need to fill in those gaps.
Oral mucositis was an orphan toxicity until about ten years ago. It has been viewed as an inevitable consequence of high-dose chemotherapy. [Clinicians have said,] ‘We know patients are going to get it; we’ll control their pain, we may have to hospitalize them, we will have to dose-reduce their chemotherapy.’ … Our goal is through a molecular-based model to work on ways to prevent or mitigate it.
As it matures, the simulation model allows us not only to generate new knowledge and identify gaps in the literature. Once it is sufficiently populated with data, and we are a ways from that … we could reach the point at which we could simulate experimental outcomes in humans before we run the Phase I, II, and III clinical trials.
How helpful is Virtual Cell when modeling mucositis?
We took the Virtual Cell software and into that [entered] calculated rate constants that were based on the software and the biological data. When I say rate constants I am talking about rate constants that were based on reaction kinetic theory. We took the data from our experimental data as well as from the literature.
What’s sophisticated about the modeling is it takes the model equations and calculates the rate constants, if you will, going forward and going backwards. It incorporates feedback loops so as, say, one pro-inflammatory change occurs, we know in humans there is going to be a feedback loop to help regulate and contain that. This model of Virtual Cell can simulate that. So it’s not a one-way pathway. It’s an integrated network.
The University of Connecticut has an NIH-funded grant for this program, Cell Analysis and Modeling. … Dr. Les Loew directs that program. We have had discussions since we first began collaborating a few years back. He deals in a nano-world of a receptor or a pathway that takes you from a messenger RNA to part of protein synthesis. …
I work in a very, if you will, macro world with cancer patients with lots of complications, lots of confounding signals. What we tried to do is move the nano-world that Virtual Cell has historically [occupied] to a cell- and tissue-based model as the next step.
Does the software lend itself to the macro world?
It sure does. We are able to predict experimental outcomes using the computational software.
[This could be] particularly [useful] when we start thinking of Phase I, II, III designs where you do your best to make an educated decision on dosing strategy for the study drug. But if [you] can run simulations, you could run thousands of simulations looking at different dosing schema for your test drug and look at the impact on kinetics and tissue before you enroll that first patient. That could add strategic value to the protocol design and perhaps avoid spending five years and $30 million running a trial that doesn’t work. I don’t want to be too hyberbolic, but this is the concept.
What were the reactions at ASCO?
No one really questions the validity and the promise of this [model]. About 40 people came by with varying backgrounds in clinical oncology, bioinformatics, biotechnology.
We wanted to be conservative, so we didn’t promise more than the model could deliver over time. … We stressed throughout the presentation that these are still very early stages in development of the model, but we would like to think it has a promising future and that it’s a novel way of thinking about mucositis research.
You used the COX-2 pathway as a prototype for this model to simulate inflammation, but the network also incorporates data from the inflammatory cascade involving COX-1, TNF-alpha, and others.
That’s because we have laboratory-based RT-PCR data on oral mucosal biopsy specimens from bone marrow transplant patients. We had biological data of associated mucositis [from] before, during, and after the transplant. … On our poster we have four what we call cubes, which are three-dimensional configurations of the histopathologic slides. Those images are directly from the microscopic slides.
One goal is, as the model matures, to link the mathematics with the experimental outcomes — that includes imaging — so that we can actually predict cellular and tissue imaging as well as clinical outcomes of this drug.
This modeling could be a tool for your research and patient care?
We are a little ways from that, but that is exactly the idea. In patients scheduled to receive high-dose head and neck radiation, say, for oral cancer, particularly if they are going to receive chemotherapy concurrently each week, we know that 100 percent of them, if I may say, are going to develop oral mucositis. About 10 percent of those persons, because they just can’t take it anymore, will take treatment breaks of the radiation. You give the tumor cells more than 72 hours to start repopulating, and that can adversely impact on tumor response to radiation and patient survival.
About 70 percent of patients undergoing high-dose chemotherapy for stem cell transplant also develop grade 3, grade 4 oral mucositis and that is where we also have the concern about myelosuppression and risk of systemic infection.
Where it is less clear is in patients with solid tumors undergoing monthly cycles of chemotherapy. … The literature would suggest anywhere from 2 [percent] to 11 [percent] to 14 percent will develop grade 3 mucositis, but we don’t know which ones up front.
If we could develop a predictive model to ask, for example, about an 11-year-old child: Is he going to be one of the nine in 10 who do not develop oral mucositis or is he going to be the one in 10 who does? That is where we think this could have real clinical value.
What does the future hold for the mucositis simulation?
We have showed the difference between the initial baseline levels and what we call the final plateau levels on some of the measures such as COX-2, the interleukins, and TNF-alpha and we show very good predictive capability by the Virtual Cell simulation versus the actual experimental data.
Where it gets more difficult to use, and we are going to be spending quite a bit of effort on this in the coming months, is in the early stages of mucosal injury; those first few weeks after the chemotherapy or head and neck radiation has started. First of all, there is much less experimental data from humans available because it is not medically safe to do biopsies or acquire tissue. There’s a gap in the literature.
In large part because of that, when we try and model what exists, we don’t get the degree of fit between the simulated and the real data versus the much better fit in the plateau level several weeks down the road. We are going to be working on all fronts, [including] the late stages, but we really want to work much more on the early stages.