Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics.
Street K, Risso D, Fletcher RB, Das D, Ngai J, Yosef N, Purdom E, Dudoit S. · 2018
Read Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. by Street K, Risso D, Fletcher RB, Das D, Ngai J, Yosef N, Purdom E, Dudoit S. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6007078.txt.
Introduction
Street and co-authors introduce Slingshot, a method for inferring cell lineages and pseudotime from single-cell data. Their paper appeared in BMC Genomics. Single-cell sequencing reveals mixed cell communities in detail. Stem cells can be charted as they mature into types. Tracing several branching paths at once stays hard. Slingshot was built for exactly that problem. It combines stable techniques suited to noisy data.
On published data sets it found one to three true trajectories, they report. Simulations showed more accurate pseudotimes than rival methods. The tool stays flexible across varied data shapes. Accurate lineage maps expose how gene activity changes over time.
Catalog record
| CREATOR | Street K, Risso D, Fletcher RB, Das D, Ngai J, Yosef N, Purdom E, Dudoit S. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics By Kelly Street, Davide Risso, Russell B Fletcher, Diya Das, John Ngai, Nir Yosef, Elizabeth Purdom, Sandrine Dudoit Published in: BMC Genomics DOI: 10.1186/s12864-018-4772-0 ABSTRACT Abstract Background Single-cell transcriptomics allows researchers to investigate complex communities of heterogeneous cells. It can be applied to stem cells and their descendants in order to chart the progression from multipotent progenitors to fully differentiated cells. While a variety of statistical and computational methods have been proposed for inferring cell lineages, the problem of accurately characterizing multiple branching lineages remains difficult to solve. Results We introduce Slingshot, a novel method for inferring cell lineages and pseudotimes from single-cell gene expression data. In previously published datasets, Slingshot correctly identifies the biological signal for one to three branching trajectories. Additionally, our simulation study shows that Slingshot infers more accurate pseudotimes than other leading methods. Conclusions Slingshot is a uniquely robust and flexible tool which combines the highly stable techniques necessary for noisy single-cell data with the ability to identify multiple trajectories. Accurate lineage inference is a critical step in the identification of dynamic temporal gene expression. Electronic supplementary material The online version of this article (10.1186/s12864-018-4772-0) contains supplementary material, which is available to authorized users. Keywords: RNA-Seq, Single cell, Lineage inference, Pseudotime inference BACKGROUND Traditional transcription assays, such as bulk microarrays and RNA sequencing (RNA-Seq), offer a bird’s-eye view of transcription. However, as they rely on RNA from a large number of cells as starting material, they are not ideal for examining heterogeneous populations of cells. Newly-developed single-cell assays can give us a much more detailed picture [1]. This higher resolution allows researchers to distinguish between closely-related populations of cells, potentially revealing functionally distinct groups with complex relationships [2]. For many systems, there are not clear distinctions between cellular states, but instead a smooth transition, where individual cells represent points along a continuum or lineage. Cells in these systems change states by undergoing gradual transcriptional changes, with progress being driven by an underlying temporal variable or pseudotime. For example, [3] examined the differentiation pattern of skeletal myoblasts, showing that their development into myocytes and mature myotubes follows a continuous lineage, rather than discrete steps. Inference of lineage structure has been referred to as pseudotemporal reconstruction and it can help us understand how cells change state and how cell fate decisions are made [3–5]. Furthermore, many systems contain multiple lineages that share a common initial state but branch and terminate at different states. These complex lineage structures require additional analysis to distinguish between cells that fall along different lineages [6–10]. Several methods have been proposed for the task of pseudotemporal reconstruction, each with their own set of strengths and assumptions. We describe a few popular approaches here; for a thorough review see [11, 12]. One of the most well-known methods is Monocle [3], which constructs a minimum spanning tree (MST) on cells in a reduced-dimensionality space created by independent component analysis (ICA) and orders cells via a PQ tree along the longest path through this tree. The direction of this path and the number of branching events are left to the user, who may examine a known set of marker genes or use time of sample collection as indications of initial and terminal cell states. The more recent Monocle 2 [8] uses a different approach, with dimensionality reduction and ordering performed by reverse graph embedding (RGE), allowing it to detect branching events in an unsupervised manner. The methods Waterfall [10] and TSCAN [7] instead determine the lineage structure by clustering cells in a low-dimensional space and drawing an MST on the cluster centers. Lineages are represented by piecewise linear paths through the tree, providing an intuitive, unsupervised method for identifying branching events. Pseudotimes are calculated by orthogonal projection onto these paths, with the identification of the direction and of the cluster of origin again left to the user. Other approaches use smooth curves to represent development, but are naturally limited to non-branching lineages. For example, Embeddr [5] uses the principal curves method of [13] to infer lineages in a low-dimensional space obtained by a Laplacian eigenmap [14]. Yet another class of methods uses robust cell-to-cell distances and a pre-specified starting cell to determine pseudotime. For instance, diffusion pseudotime (DPT) [6] uses a weighted k nearest neighbors (kNN) graph on cells and calculates distances using transition probabilities over random walks of arbitrary length. Similarly, Wishbone [9], an extension of Wanderlust [4], uses an ensemble of kNN graphs on cells along with a randomly selected group of waypoints to iteratively refine stable distance estimates. Finally, other methods take a model-based approach to detecting branching events. GPfates [15] uses a Gaussian process latent variable model (GPLVM) and overlapping mixtures of Gaussian processes (OMGP) to infer trajectories and pseudotimes. A similar method, DeLorean [16], uses a single GPLVM to infer pseudotimes along a single trajectory. And the mixtures of factor analysers (MFA) method [17] takes a hierarchical Bayesian approach, using Markov chain Monte-Carlo (MCMC) to sample from the posterior of a fully generative model that includes branch identities. See Table 1 for a summary of existing methods. [Table: Summary of existing lineage and pseudotime inference methods] Here, we introduce Slingshot, a novel lineage inference tool designed for multiple branching lineages. Slingshot combines highly stable techniques necessary for noisy single-cell data with the flexibility to identify multiple lineages with varying levels of supervision. Slingshot consists of two main stages: 1) the inference of the global lineage structure and 2) the inference of pseudotime variables for cells along each lineage (Fig. 1). Like other methods [7, 10], Slingshot’s first stage uses a cluster-based MST to stably identify the key elements of the global lineage structure, i.e., the number of lineages and where they branch (Fig. 1, Step 1). This allows us to identify novel lineages while also accommodating the use of domain-specific knowledge to supervise parts of the tree (e.g., terminal cellular states). For the second stage, we propose a novel method called simultaneous principal curves, to fit smooth branching curves to these lineages, thereby translating the knowledge of global lineage structure into stable estimates of the underlying cell-level pseudotime variable for each lineage (Fig. 1, Step 2). The Slingshot method is implemented in the open-source R package slingshot (available from the GitHub repository https://github.com/kstreet13/slingshot) to be released through the Bioconductor Project (http://www.bioconductor.org). [Figure: Schematics of Slingshot’s main steps. The main steps for Slingshot are shown for: Panel (a) a simple simulated two-lineage two-dimensional dataset and Panel (b) the single-cell RNA-Seq olfactory epithelium three-lineage dataset of [26] (see Results and discussion for details on dataset and its analysis). Step 0: Slingshot starts from clustered data in a low-dimensional space (cluster labels indicated by color). For Panel (b), the plot shows the top three principal components, but Slingshot was run on the top five. Step 1: A minimum spanning tree is constructed on the clusters to determine the number and rough shape of lineages. For Panel (b), we impose some constraints on the MST based on known biology. Step 2: Simultaneous principal curves are used to obtain smooth representations of each lineage. Step 3: Pseudotime values are obtained by orthogonal projection onto the curves (only shown