Quantifying of Stem Cell Toxicology with Deep Neural Network

The high volume of complex data collected during stem cell toxicology experiments necessitates the use of computational analysis tools. Modalities of data including microscopy images and molecular data such as DNA/RNA sequences, and protein information provide unique insights into cellular physiology. For example, behavioral dynamics of colonies are observed via time-lapse microscopy images, and analysis of proteins on the surface of, and within cells, provides a more detailed explanation of the underlying mechanisms of these changes. The goal of this project is to use Deep Learning, a subset of Artificial Intelligence computer algorithms, to combine microscopy and molecular data into a single, predictive model of stem cell development under toxic exposure to chemicals from cigarettes and cigarette alternatives. Data will be collected, over time, in the aforementioned modalities, for experiments involving exposure of human embryonic stem cell colonies to various combinations of cigarette chemicals. The database established from these experiments will be used to design and test multiple configurations of our model in tasks related to determining temporal mechanisms of cell health and other developmental changes. The results of the experiments will be disseminated in a variety of manners including peer reviewed journals, and community engagement, in order to inform the public of the developmental risks associated with the use of tobacco products.