| Stem cells are functionally defined cells with a high therapeutic potential for many diseases. The stemness hypothesis states that stem cells share a core set of mechanisms that regulate the shared stem cell properties of self-renewal and multi-lineage potential. Previous attempts to identify genes required for core stem cell function across stem cell types using transcriptional profiling have identified few such genes. My work focused on the development of a computational stemness meta-analysis (SMA) method that uses high-throughput differential gene expression data integration to address three main questions: do functional redundancy and tissue-specific expression mask common molecular mechanisms shared between stem cell types? Are stemness mechanisms conserved between mouse and human stem cells? Can we use gene expression signatures to predict stem cell state?;The SMA method identified 103 mouse evolutionarily related groups of homologous genes with reproducible, statistically significant, cell type diverse and stem cell-specific upregulation in multiple stem cell types. The results point to specific examples of functional redundancy in modules controlling cell adhesion, quiescence, and gene silencing. Shared homolog modules also include genes in the Myc, Myb, CM, Hspa, Id, and many other families. Genes within the stemness homolog families are prime candidate regulators of conserved stemness mechanisms and may play critical roles as stem cell markers.;I directly measured the level of conservation of stemness mechanisms between mouse and human cells. Application of the SMA method to a human stem cell compendium indicates that human data are globally more heterogeneous than murine stem cell data. However, human stemness families incorporate several conserved mammalian stemness modules, such as the Integrin aalpha, TCF/LEF, Frizzled, Notch, and Chd families.;Finally, I used the stemness modules identified in the mouse SMA to define a stemness index score and evaluate how stem cell-like a new gene expression signature is. I validated the predictiveness of the stemness modules through an internal cross-validation test and applied the stemness index test to a large set of new experiments from normal stem cells, side populations, cancer stem cells, and metastatic populations. The results indicate that mouse stemness modules could predict stem cell-like features in various data sources with high accuracy. |
