CellRank: consistent and data view agnostic fate mapping for single-cell genomics

Single-cell RNA sequencing quantifies biological samples at an unprecedented scale, allowing us to decipher biological differentiation dynamics such as normal development or disease progression. As conventional single-cell RNA sequencing experiments are destructive by nature, reconstructing cellular trajectories computationally is an essential aspect of analysis pipelines. To infer trajectories in a consistent and scalable manner, we have developed CellRank. In its first iteration, CellRank quantitatively recovered trajectories from RNA velocity estimates and transcriptomic similarity. Given these data views, CellRank constructed a cell-cell transition matrix, inducing a Markov chain to automatically infer terminal states and describe their lineage formation. However, CellRank did not enable incorporating complementary data views such as experimental time points, pseudotime or stemness potential. To facilitate these and future views, CellRank 2 generalizes CellRank's trajectory inference framework to multiview single-cell data, leading to a general and scalable framework for cellular fate mapping. Overall, the CellRank framework enables the consistent quantification of cellular fate, combining complementary views and analyzing lineage priming consistently. Here we provide detailed protocols on how to run exemplary CellRank analyses at scale and across different data views. Using CellRank requires basic apprehension and knowledge of single-cell omics data and the Python programming language.