TABLE 2.

Computational approaches utilized by scRNA-seq studies profiling the bone marrow.

Operation	Tool/Algorithm	Description
Normalization, batch correction, clustering	Seurat (Butler et al., 2018; Stuart et al., 2019)	An R package for quality filtering, normalization, dimensionality reduction, and visualization of scRNA-seq data. It additionally includes a method for integrated analysis of multiple datasets by identifying pairwise correspondences between single cells across those datasets.
Visualization	t-SNE (van der Maaten and Hinton, 2008)	Non-linear dimensionality reduction technique based on Student-t distribution for converting data in a high-dimensional space to a low-dimensional one while avoiding overcrowding.
	SPRING (Weinreb et al., 2018a)	Force-directed graph layout for visualizing continuous topologies that generates a graph of nodes representing cells connected to their nearest neighbors in high-dimensional gene expression space.
	UMAP (Becht et al., 2018; McInnes et al., 2018)	Approximates a manifold and a constructs its fuzzy topological representation with the goal of preserving more of the global structure.
Pseudotime or trajectory inference	DPT/Destiny (Haghverdi et al., 2015; Angerer et al., 2016)	Uses diffusion maps to identify the low-dimensional structure, then identifies a pseudotime metric based on transition probabilities of differentiating toward various cell fates.
	Monocle (Trapnell et al., 2014; Qiu et al., 2017)	Constructs a minimum-spanning tree (MST) through the dimension-reduced space created by independent component analysis. Cells are ordered along the longest path through the MST. Monocle was later modified to use DDRTree for dimensionality reduction and ordering.
	PAGA (Wolf et al., 2019)	A partition-based graph abstraction tool that provides a coarse-grained representation by placing edges between cluster nodes with similar cells. Unlike many trajectory inference methods, it is able to account for disconnected topologies.
	PBA (Tusi et al., 2018; Weinreb et al., 2018b)	Uses nearest neighbor graph cell densities to predict fate probabilities and the direction of differentiation.
	Slingshot (Street et al., 2018)	Uses a cluster-based MST to identify the lineages and where they branch, then uses simultaneous principal curves to fit a smooth representations of each lineage.
	STEMNET (Velten et al., 2017)	Uses hierarchical clustering to define the most differentiated cell populations and then uses those populations as a training set for classifying priming in the less mature populations.
Velocity	Velocyto (La Manno et al., 2018)	Predicts the future state of single cells based on the relative abundance of unspliced precursor and spliced mature mRNA.
Cell-cell interactions	RNA-Magnet (Baccin et al., 2020)	Estimates the likelihood of physical interactions between single cells based on the expression of cell-surface receptors and their binding partners.