Supervised classification enables rapid annotation of cell atlases

By Hannah A. Pliner, Jay Shendure, Cole Trapnell. Nature Methods, 2019

426 citations recorded by the source corpus.

Single-cell molecular profiling technologies are gaining rapid traction, but the manual process by which resulting cell types are typically annotated is labor intensive and rate-limiting. We describe Garnett, a tool for rapidly annotating cell types in single-cell transcriptional profiling and single-cell chromatin accessibility datasets, based on an interpretable, hierarchical markup language of cell type-specific genes. Garnett successfully classifies cell types in tissue and whole organism datasets, as well as across species. Garnett uses a hierarchical markup language and machine learning to define cell types and their marker genes and identifies these cell types in scRNA-seq datasets from tissues and whole organisms and across species.

Subfields: Biophysics, Cancer Research, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/5564831ff39edbea38d3cef198da5e565cc5475f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biophysics

1. Aleksandr Ianevski
2. Anil K. Giri
3. Tero Aittokallio
4. Cole Trapnell

In Cancer Research

1. Tero Aittokallio
2. Aleksandr Ianevski
3. Anil K. Giri
4. Cole Trapnell
5. Jay Shendure
6. Jason D. Buenrostro

In Molecular Biology

1. Caleb A. Lareau
2. Cole Trapnell
3. Tero Aittokallio
4. Aleksandr Ianevski
5. Anil K. Giri
6. Jay Shendure
7. Hannah A. Pliner
8. Frank Steemers

Citations in the corpus

• Assessment of computational methods for the analysis of single-cell ATAC-seq data (2019, methodology): “547 Based on recent studies [9, 19], we expected ~8 populations: CD34+, Natural Killer and 548 Dendritic cells, Monocytes, lymphocyte B and lymphocyte T cells, together with 549 terminally differentiated CD4 and CD8 cells.”
• Fully-automated and ultra-fast cell-type identification using specific marker combinations from single-cell transcriptomic data (2022, methodology): “…study, by segregating them into immature B-cell and plasma (B) cell types (lower UMAP plot of panel ( d ). that was recently shown to outperform several popular tools such as Garnett 31 and CellAssign 32 ), (ii) fastest running time (SCINA 33 ), and (iii) full-automated process (scCATCH 34 ).”
• A human cell atlas of fetal gene expression (2020, methodology): “In addition to these manual annotations of cell types, we also generated semiautomated classifiers for each organ using Garnett (99).”
• scDeepSort: a pre-trained cell-type annotation method for single-cell transcriptomics using deep learning with a weighted graph neural network (2021, methodology): “As a consequence, the poor performance of CellAssign, Garnett, and SCINA might result from the poor extrapolation of markers on cell-based marker-dependent cell typing meth-ods.”