|
Abstract:
|
New technologies enabling genomic-scale profiling of single cells offer enormous promise to identify novel cell types and states, reconstruct fine-scale regulatory networks, and deconvolute heterogeneous responses in complex systems. However, the high level of technical and biological noise that is inherent in single cell data poses significant analytical challenges. I will present two machine learning based algorithms for analyzing single cell data, and apply them to interpret cellular heterogeneity in human blood, the zebrafish embryo, and the mouse retina. The first method combines linear and non-linear dimensional reduction techniques for unbiased discovery of cell types based on gene expression data. The second approach combines single cell transcriptomics with in situ hybridization in order to learn a cell's spatial origin from within a complex tissue. Finally, I will discuss a novel technological approach, leveraging droplet microfluidics, to rapidly and cheaply prepare tens of thousands of single cell libraries.
|
