Cell identity is intricately shaped by the control of gene expression. In our primer and seminar,
we explore current methods to define and control cell identity. We will present our novel
Method, CellOracle, a machine learning-based tool, that utilizes data from single-cell multi-omics to explore how transcription factors (TFs) regulate cell identity. We will begin with a broad, conceptual overview of CellOracle – highlighting how the design of the method enables insights into how cell identity shifts following in silico TF perturbation. We will cover the application of CellOracle to well-established biological systems such as mouse and human hematopoiesis, as well as zebrafish embryogenesis. We will demonstrate how the framework accurately models phenotypic changes resulting from TF perturbations, validating its predictive capabilities against known biological phenomena. Taking a deeper dive into application, we will highlight the systematic in silico perturbation of TFs in the developing zebrafish, leading to novel insights into the well-studied process of axial mesoderm formation. We will also explore the application of CellOracle to cellular reprogramming and cover the limitations of the approach. Building on this overview of the method and its application.