CellRank is a computational framework to study cellular fate decisions based on various types of single-cell genomics data. CellRank scales to large cell numbers, is fully compatible with the scverse ecosystem, and is easy to use.
Single-cell genomics technologies record cellular heterogeneity at ever-increasing scale and resolution. However, they remain cell-destructive and require computational trajectory inference to reconstruct cellular dynamics. Prior work used gene expression similarity to estimate pseudotime, a simplified one-dimensional cell ordering. While this concept was successfully applied to well-characterized processes, it remains challenging to reconstruct trajectories for complex biological systems due to uncertain directions and complex dynamics.
With CellRank, we extend trajectory inference beyond well-characterized systems to gain insights into disease-relevant systems, including development, regeneration, reprogramming, and cancer. We achieve this by exploiting additional data modalities, like RNA velocity (La Manno et al., Nature 2018; Bergen et al., Nature Biotech 2020), metabolic labeling (Erhard et al., Nature 2019; Battich et al., Science 2020), spatial context, experimental time points (Schiebinger et al., Cell 2019), and lineage tracing information (Wagner et al., Nat. Rev. Gen., 2020).
The project operates under the BSD-3-Clause license, promoting open redistribution and usage with or without modifications.
Detailed guidance and information are available in the official documentation, facilitating a deeper understanding of CellRank's applications and functionalities.
We demonstrate the success of our software in two publications:
The main contributors to the CellRank project are the members of the Theis Lab. You can find more about the team and their contributions on the GitHub page.
Important biological processes, including development, regeneration, and cancer, manifest as complex cellular state changes. Studying these processes holds enormous potential for many areas of medicine. Single-cell genomics (SCG) represents a promising toolset to gain insights into such processes as it can resolve individual molecular states. For example, single-cell RNA-sequencing (scRNA-seq) measures gene expression in single cells. However, scRNA-seq and related assays are destructive, representing a hurdle to link fate outcomes to molecular state changes. Thus, the main task of CellRank is to gain biological insights into dynamical cellular processes from SCG data.
Concrete challenges faced by prior methods include:
CellRank overcomes these challenges by (1) operating in latent spaces and aggregating over many cellular profiles in Markov chains, (2) estimating directionality from various experimental priors, (3) interfacing with optimized linear algebra libraries, and exploiting sparsity.
For bugs, help, or suggestions, feel free to open an issue or reach out via email at info@cellrank.org.
Should you have any specific inquiries regarding CellRank, please do not hesitate to reach out to Philipp Weiler at philipp.weiler@helmholtz-munich.de or to Marius Lange at marius.lange@helmholtz-munich.de
CellRank is an open-source software package that is already used by biologists and bioinformaticians around the world to analyze complex cellular dynamics in situations like cancer, reprogramming or regeneration.
CellRank, which Pe'er also had a hand in, came out last year to assist in single-cell fate mapping. Just last week, lead CellRank developer Fabian Theis and colleagues at the Institute of Computational Biology at Helmholtz Munich in Neuherberg, Germany, unveiled a method called CellRank 2 in a preprint posted to BioRxiv. This update enables the study of cellular fate with large-scale single-cell data.
Following our past coverage of cellular fate mapping, a new pre-print details the launch of CellRank2, a versatile and scalable framework for studying cellular fate using multiview single-cell data.
Scanpy is a scalable toolkit for analyzing single-cell gene expression data built jointly with anndata. It includes preprocessing, visualization, clustering, and differential expression testing. The Python-based implementation efficiently deals with datasets of more than one million cells.