Cells differentiate at different speeds: even
when collected at the same time point, cells may be at different stages
of the differentiation process, exhibiting different gene expression
profiles. Additionally, single cells are often a mixture of
multiple subtypes, sometimes including previous unknown subtypes and
subtypes that were sampled due to imperfection of the experimental
procedure. We aim to infer the cell differentiation trajectory
from single-cell gene expression data, while accounting for potentially
multiple paths and possibly unknown and unintended subtypes. We
cluster cells by their gene expression profiles into `super cells’ and
infer a causal graph among the super cells. The causal graph
corresponds to the differentiation process at a coarse level.
Individual cells are placed along this process depending on their
distance from the nearest super cells. I will illustrate this
idea with examples and explain the current development of methods and
algorithms for this inference.
|