Decoding the cornea: Computational approaches for analysis of cell identity in single cell genomics data

This thesis investigates corneal cell identity at single-cell resolution across donor-derived tissue, pluripotent stem cell-derived corneal cells, and mouse models of corneal injury. By integrating and analyzing single-cell RNA-sequencing and single-cell ATAC-sequencing datasets, the work advances understanding of corneal cell heterogeneity, differentiation, and regeneration. An integrated human corneal atlas compiled from multiple public datasets redefines corneal cell identity, improves marker gene identification, and reveals previously unrecognized cell populations. The atlas also supports a machine learning framework for cell classification and enables identification of transcription factors regulating corneal cell states. Using pluripotent stem cell differentiation models, this thesis demonstrates variability in corneal stem cell generation between cell lines and identifies molecular markers that distinguish desired corneal stem cells from unwanted cell populations. Multi-omic integration further reveals transcriptional regulators associated with successful differentiation and persistent pluripotency. In a mouse corneal injury model, single-cell analysis shows that sutured tissue exhibits features of corneal stem cell deficiency, while Duloxetine treatment reduces inflammation and alters signaling pathways. Finally, a novel computational framework, scANANSE, is introduced to infer transcription factors driving differences between cell states. Together, this thesis demonstrates the importance of single-cell and multi-omic approaches for understanding corneal biology and improving regenerative strategies.