可以看到bioRxiv上是November 02, 2018发布的，然后Cell主刊June 06, 2019正式发表。
Single-cell transcriptomics has transformed our ability to characterize cell states, but deep biological understanding requires more than a taxonomic listing of clusters.
As new methods arise to measure distinct cellular modalities, a key analytical challenge is to integrate these datasets to better understand cellular identity and function.
Here, we develop a strategy to “anchor” diverse datasets together, enabling us to integrate single-cell measurements not only across scRNA-seq technologies, but also across different modalities.
After demonstrating improvement over existing methods for integrating scRNA-seq data, we anchor scRNA-seq experiments with scATAC-seq to explore chromatin differences in closely related interneuron subsets and project protein expression measurements onto a bone marrow atlas to characterize lymphocyte populations.
Lastly, we harmonize in situ gene expression and scRNA-seq datasets, allowing transcriptome-wide imputation of spatial gene expression patterns.
Our work presents a strategy for the assembly of harmonized references and transfer of information across datasets.
- immunophenotype (Stoeckius et al., 2017; Peterson et al., 2017),
- genome sequence (Navin et al., 2011; Vitak et al., 2017),
- lineage origins (Raj et al., 2018; Spanjaard et al., 2018; Alemany et al., 2018),
- DNA methylation landscape (Luo et al., 2018; Kelsey et al., 2017),
- chromatin accessibility (Cao et al., 2018; Lake et al., 2018; Preissl et al., 2018),
- spatial positioning
- how can disparate single-cell datasets, produced across individuals, technologies, and modalities be harmonized into a single reference
- once a reference has been constructed, how can its data and meta-data improve the analysis of new experiments?
These questions are well suited to established fields in statistical learning.
第二个问题就类似reference assembly (Li et al., 2010) and mapping (Langmead et al., 2009) for genomic DNA sequences
identify shared subpopulations across datasets
- canonical correlation analysis (CCA)
- mutual nearest neighbors (MNNs)
- only a subset of cell types are shared across datasets
- significant technical variation masks shared biological signal.
- reference assembly
- transfer learning for transcriptomic, epigenomic, proteomic,
- spatially resolved single-cell data
Through the identification of cell pairwise correspondences between single cells across datasets, termed ‘‘anchors,’’ we can transformdatasets into a shared space, even in the presence of extensive technical and/or biological differences.
This enables the construction of harmonized atlases at the tissue or organismal scale, as well as effective transfer of discrete or continuous data from a reference onto a query dataset.
false negatives (‘‘drop-outs’’) due to transcript abundance and protocol-specific biases
expression derived from fluorescence in situ hybridization (FISH) exhibits probe-specific noise due to sequence specificity and background binding
Identifying Anchor Correspondences across Single-Cell Datasets
基本的假设：we assume that there are correspondences between datasets and that at least a subset of cells represent a shared biological state.
Constructing Integrated Atlases at the Scale of Organs and Organisms
Leveraging Anchor Correspondences to Classify Cell States
Projecting Cellular States across Modalities
Transferring Continuous and Multimodal Data across Experiments
Predicting Protein Expression in Human Bone Marrow Cells
Spatial Mapping of Single-Cell Sequencing Data in the Mouse Cortex
what's my problem?