"Characterizing transcriptional cell-to-cell heterogeneities during differentiation"
Location: Lecutre hall in Biocentre (B01.027)
Cell-to-cell variations in gene expression underlie many biological processes. Currently more and more experimental tools are becoming available in order to observe these variations, and to draw conclusions on underlying processes. However, given these experimental advances, we are now facing a series on computational questions dealing with these data, since classical analysis tools are often tailored to population averages.
Here I will discuss the statistical analysis and network modeling of single-cell qPCR expressions using nonlinear dimension reduction applied to stem cell differentiation. The analysis is based on a recently proposed framework based on Gaussian process latent variable models, which we extend to guarantee that small distances are preserved for cells from the same developmental stage [Buettner & Theis; Bioinformatics 2012]. Based on our recent characterization of transcriptional networks during blood stem cell differentiation [Moignard et al, Nat Cell Biol 2013], we show how to identify subpopulations using mixture models, extended in order to allow for censoring inherent to the qPCR measurements. We then are able to recapitulate the differentiation hierarchy by quantifying cell-to-cell distances in the transcript space using diffusion map embedding.
In summary, we show that single-cell expression allows for a robust and far more detailed description of differentiation decisions than using population averages.
You can find more information about Professor Theis and his research on the website of the Helmholtz Centre Munich.