Statistical inference for differential analysis in single cell transcriptomics

The field of single cell gene expression analysis (scRNA-seq) is experiencing an unprecedented boom
in popularity. It enables researchers to study genes in individual cells at an unpreceded resolution
and will revolutionise their view on complex biological processes, tissues and disease. The current
technology, for instance, can profile the expression of thousands of genes in up to ten thousand
cells of a sample. This can give scientists a bird's-eye view on the gene expression profiles of all
different cell types in a sample and it has the promise to unravel each of their driver genes or gene
signatures, e.g. for the various white, red and platelet cell types in a blood sample. More
importantly, it can also establish how these signatures differ between healthy and diseased
subjects; and how each cell type responds to a treatment. The discovery of gene signatures and
their differences upon stimulation, however, is currently hampered by the lack of data analysis tools
that 1) can deal with many genes that go missing because protocols start from the tiny amount of
input material of a single cell, 2) scale to the huge data volumes in scRNA-seq experiments, and 3)
account for strong correlations as multiple cells within a subject typically have expression profiles
that are more alike than those of different subjects. Within this project we will therefore develop
novel statistical tools that overcome each of these challenges, thus greatly enhancing scRNA-seq
data analysis.