Chase Sakitis, PhD, Post Doctoral Research Scholar, Health Services & Outcomes Research, received a TL1 Post-Doctoral Trainee Award from National Institutes of Health (NIH) National Center for Advancing Translational Sciences (NCATS) via a Frontiers CTSI subaward.
The funding is being used for Dr. Sakitis’ project, “Bayesian models for the analysis of spatial proteomic data with application to ovarian cancer”. It covers a project period of July 1, 2025-June 30, 2027, with an estimated $81,000 for the first year of the award.
As Dr. Sakitis explains, development of immunotherapy (IO) has been groundbreaking in cancer treatment and has shown promise in treating melanoma, lung cancer, and renal cell carcinoma. However, this has not been effective in other cancers. Clinical trials investigating IO often show little to no signs of immune activity. This makes it difficult to determine characteristics in the tumor immune microenvironment (TIME) that might be associated with IO response. This highlights the importance of tumor immune profiling to identify predictive markers of therapeutic response.
Multiplex immunofluorescence (mIF) is a developed state-of-the-art technology that has been used for immune profiling. It involves a lab technique where scientists use glowing tags that stick to specific proteins in cells, allowing scientists to see where the proteins are. Multiplex immunofluorescence helps scientists and doctors better understand diseases like cancer, see how immune cells are reacting to a tumor, and develop better treatments by knowing which cells are active.
Even though multiplex immunofluorescence is a powerful tool, it comes with some problems when scientists try to analyze the data. Due to the low immune activity, like with ovarian cancer, the data experiences high variation, zero-inflation (no positive cell counts), and multiple samples from the same subject. Standard statistical models can be utilized to model the data but does not necessarily account for all potential characteristics of the data.
In this research, Dr. Sakitis proposes to overcome these challenges through the development of novel Bayesian statistical models and software.
Bayesian models are used when you have limited data, you want to combine past knowledge with new information, and you’re dealing with uncertainty that should be expressed clearly. These models incorporate potential relationships between features of interest which the standard models do not consider.
The effectiveness of the novel Bayesian models will be tested in three large ovarian cancer studies, where mIF data has been collected on multiple markers (e.g., CD3, CD8, CD69, FoxP3).
“The ability to apply these new methods to study the tumor immune microenvironment of ovarian cancer will allow for the development of an ovarian cancer prognostic immune signature that could help with treatment decisions,” explains Dr. Sakitis.
Brooke Fridley, PhD, Director, Biostatistics & Computational Biology Core, Division Director, Health Services & Outcomes Research, serves as a mentor.