Machine-guided cell-fate engineering

The creation of induced pluripotent stem cells (iPSCs) has enabled scientists to explore the function, mechanisms, and differentiation processes of many types of cells. One of the fastest and most efficient approaches is transcription factor (TF) over-expression. However, finding the right combination of TFs to over-express to differentiate iPSCs directly into other cell types is a difficult task. Here, we describe a machine-learning (ML) pipeline, called CellCartographer, that uses chromatin accessibility and transcriptomics data to design multiplex TF pooled-screening experiments for cell-type conversions that then may be iteratively refined. We validate this method by differentiating iPSCs into twelve cell types at low efficiency in preliminary screens and iteratively refine our TF combinations to achieve high-efficiency differentiation for six of these cell types in <6 days. Finally, we functionally characterize iPSC-derived cytotoxic T cells (iCytoTs), regulatory T cells (iTregs), type II astrocytes (iAstIIs), and hepatocytes (iHeps) to validate functionally accurate differentiation.