Intracellular Heterogeneity: Seeing the Branches through Gaussian Mixture Modeling

Prior to comprehensive genetic profiling techniques, cancer-specific tumors were expected to be similar regardless of the patient; now, it is known that cancer tumors are in fact quite diverse and each tumor consists of diverse subpopulations. Common approaches to characterizing a cancer population include recovering the distribution of the average light intensity of channels per biopsy image; unfortunately, such approaches insufficiently characterize the biopsy. However, through genetic testing and in situ imaging of tissue sections via fluorescence-based immunohistochemistry, it is possible to analyze the spatial distribution of specific biomarkers within a single cancer biopsy. In this work, the intracellular heterogeneity (ICH) spatial distribution was analyzed via Gaussian mixture models and revealed the number of characteristics, or branches, required to suitably describe the distribution. Essentially, ICH spatial pattern recognition in cancer biopsies may aid clinicians’ diagnosis and prognosis, and perhaps ease the development of personalized cancer treatments.