Multiple neural network approaches, including use of topological data analysis, enhances classification of human induced pluripotent stem cell colonies by treatment condition

Abstract Understanding how stem cells organize to form early tissue layers remains an important open question in developmental biology. Helpful in understanding this process are biomarkers or features that signal when a significant transition or decision occurs. We show such features from the spatial layout of the cells in a colony are sufficient to train neural networks to classify stem cell colonies according to differentiation protocol treatments each colony has received. We use topological data analysis to derive input information about the cells’ positions to a four-layer feedforward neural network. We find that despite the simplicity of this approach, such a network has performance similar to the traditional image classifier ResNet. We also find that network performance may reveal the time window during which differentiation occurs across multiple conditions. Author summary Our understanding of how stem cells determine what specialized cells to differentiate into is still incomplete. One aspect of understanding this process involves identifying when key decisions about a cell’s fate occur. We explored whether by looking at the layout of the cells of the colony, we can infer knowledge regarding the eventual phenotypes the cells are differentiating towards. We train an algorithm to recognize cell type using spatial information by taking as input the number and size of holes that appear among the colony’s cells. We find this method succeeds in its classification, similar to an industry-grade image classifier. Competing Interest Statement The authors have declared no competing interest.