Moirai: single-cell trajectory inference grounded in gene-level expression dynamics

Cell differentiation is a crucial process in the development of multicellular organisms, and it is regulated by a series of coordinated changes in gene expression. Various methods exist to utilize single-cell RNA-seq data to infer a cell's position along a pseudo-temporal axis and identify significant genes involved in this process.

These trajectory inference methods usually focus on global transcriptomic changes and mathematical algorithms. However, concentrating too much on these broad changes might reduce the method's sensitivity to detect branching points and convergent trajectories, which are often determined by minute-scale transcriptional events.

Recognizing this limitation, a team of researchers created Moirai, a graph-based trajectory inference method. This novel method identifies dynamic gene expression patterns throughout a developmental continuum and utilizes these patterns to establish a common pseudotime axis for all cells. By shifting the focus to individual gene dynamics, Moirai enhances its ability to identify potential branching points that may be obscured by global transcriptomic similarities.

The researchers tested Moirai on four developmental datasets and found that it successfully recovers gene expression patterns of genes known to be involved in the specific developmental process. This evidence supports the idea of using these genes to define a cell's pseudotime. Moreover, Moirai demonstrated the ability to robustly infer gene expression patterns across various embedding approaches, underscoring the value of prioritizing small-scale transcriptional dynamics in the inference process.