Despite scRNA-Seq data provides a snapshot of gene expression at a particular time point, expression levels are captured across individual cells. It captures the heterogeneity of cells within a population. If biological tissue was not a steady state, it may be possible to retrieve trajectories, such as developmental lineage or disease progression. Trajectory inferences and RNA velocity leverage the information within that snapshot to infer such cellular trajectories. Ordering cells is a first step to understand the underlying biological processes driving cell fate decisions and differentiation events.

Abbreviations :

TI : trajectory inference
kNN : k-nearest neighbors

Method principles

In section, we make a overview of two main methods to order cells : trajectory inference and RNA velocity.

Trajectory inference

Trajectory inference is based on the gene expression profile. It looks at distance between cells : cells with similar transcriptomic profile are close, and cells with distinct profile are distant. Based on transcriptomic profiles, it finds paths between cells, leading to trajectories.


Figure from Wolf et al. (2019). Partition-based graph abstraction generates a topology-preserving map of single cells. High-dimensional gene expression data is represented as a kNN graph by choosing a suitable low-dimensional representation and an associated distance metric for computing neighborhood relations. […] We use PCA-based representations and Euclidean distance. […] By discarding spurious edges with low weights, PAGA graphs reveal the denoised topology of the data at a chosen resolution and reveal its connected and disconnected regions. […] We order cells within each partition according to their distance from a root cell. A PAGA path then averages all single-cell paths that pass through the corresponding groups of cells. This allows to trace gene expression changes along complex trajectories at single-cell resolution

There are more than 50 methods to perform trajectory inference because : - many input types are possible (count matrix, normalized matrix, reduced space) - many ways to compute distance - many ways to find a path

How to choose a trajectory inference method ?