26.07.2021

Single Cell RNAseq Analysis Workflow

Trajectory Inference

  • Cells that differentiate display a continuous spectrum of states

    • Transcriptional program for activation and differentiation
  • Individual cells will differentiate in an unsynchronized manner

    • Each cell is a snapshot of differentiation time
  • Pseudotime – abstract unit of progress

    • Distance between a cell and the start of the trajectory

Should you run trajectory inference?

  • Are you sure that you have a developmental trajectory?

  • Do you have intermediate states?

  • Do you believe that you have branching in your trajectory?

  • Be aware, any dataset can be forced into a trajectory without any biological meaning!

  • First make sure that gene set and dimensionality reduction captures what you expect.

Trajectory Inference Methods

  • Saelens et al. (2019) Nature Biotechnology

Which method to use?

Diffusion Maps (in brief)

  • It is a NON-LINEAR method of dimensionality reduction.

  • The distances between points are measured as probability from going from one to another.

  • The data must present connectivity (transitional cells).

Final Considerations

  • In reality, distance in multidimensional space reflects difference in transcriptional landscape, not actual time.

  • Necessary to have a continuum of states among your cells (Will not work well with 2 distinct clusters)

  • May work with single time-point if ongoing differentiation process (It is better to have multiple experimental time points)

  • Be aware, any dataset can be forced into a trajectory without any biological meaning!