Machine Learning Strategy to Determine Cell Shape Phenotypes Associated with Micro-Environmental Cues
Desu Chen, Sumona Sarkar, Julian Candia, Carl G. Simon Jr., Wolfgang Losert, Joy P. Dunkers, Stephen J. Florczyk, Subhadip Bodhak, Meghan Driscoll
Cell morphology has been identified as a potential indicator of stem cell response and function in biomaterial environments. However, determination of cell shape phenotype in biomaterials is complicated by heterogeneous cell populations, heterogeneous microenvironments, and multi-parametric definitions of cell morphology. To associate cell morphology with cell-material interactions, we developed an analysis framework based on support vector machines (SVMs) to build classifier boundaries that describe and predict microenvironment-driven cell morphology differences. Due to high heterogeneity in cell morphology at the single-cell level, a supercell averaging method was implemented to improve classification performance and robustness of SVMs. We compared morphologies of human bone marrow stromal cells (hBMSCs) cultured on nanofiber scaffolds to those on flat films in the presence or absence of osteogenic differentiation media after one day of culture. Smaller cell size and more dendritic shape patterns were the major morphological responses of hBMSCs to nanofiber scaffolds. Upon osteogenic supplement treatment, an increase in concavity and boundary roughness of hBMSCs was identified. hBMSC morphology was more strongly affected by nanofiber topologies than by chemical stimulation. This analysis technique will allow the identification of stem cell morphological features that may be associated with critical biological functions and enable targeted biomaterial design.