Fig. 1: CellRank 2 provides a unified framework for studying single-cell fate decisions using Markov chains.
From: CellRank 2: unified fate mapping in multiview single-cell data
CellRank 2 uses a modular design. Data and problem-specific kernels calculate a cell–cell transition matrix inducing a Markov chain (MC); of these kernels, at least one has to be used, but multiple can be combined via a kernel combination (Methods). Estimators analyze the MC to infer initial and terminal states, fate probabilities and lineage-correlated genes. Using fate probabilities and a pseudotime allows for studying gene expression changes during lineage priming. Features inherited from the original CellRank implementation are indicated in blue and new features are in orange.
