Single cell lineage tracing in transcriptomics (TBC)

2018 review "Single-Cell Transcriptomics Meets Lineage Tracing", cell stem cell

Dimensionality reduction-based algorithms

Monocle, independent-component analysis(ICA), project all cells in 2 dimensions, followed by minimum spanning trees(MST) construction to form a backbone to connect the most and least differentiated cells. All remaining cells are projected on the backbone, resulting in a 1-dimensional ordering of all cells, which is pseudotime.

disease model: developing human skeletal muscle myoblasts.

limitation: inability to allow for bifurcation in the lineage prediction, only linear differentiation systems can be analyzed. (which be solved in monocle2)

Monocle2, builds the lineage tree in a higher dimensional space.

disease model: several types of neurons, progenitors and stem cells of the hematopoietic system

Slingshot, similar strategy as Monocle by 1. reducing dimensionality of the data via PCA(principal component analysis), ICA, or t-stochastic neighbor embedding (tSNE) and then constructing a MST(minimum spanning tree) .

Nearest Neighbor Graph-Based Algorithms

based on k-NNG(k-nearest neighbor graphs)

PAGA

Monocle3

Other Lineage Reconstruction Algorithms

using k-medoid or hierarchical clustering

RNA velocity

the current state of the cell is transcripts only containing exons 

fraction of unspliced transcripts across all genes defines as RNA velocity vectors, is the future state of the cell.

both the current state and the future state can then be projected into a 2-dimensional space and show the direction and the speed of differentiation.

Wet-lab technique

GFP (green fluorescent proteins)

DNA barcodes, viral strategies, extend to include cell-specific DNA barcodes.