Single-cell assignment using multiple-adversarial domain adaptation network with large-scale references

Cell Reports Methods, 2023 · DOI: https://doi.org/10.1016/j.crmeth.2023.100577 · Published: September 25, 2023

Simple Explanation

This paper introduces SELINA (single-cell identity navigator), an integrative and automatic cell-type annotation framework. SELINA employs a multiple-adversarial domain adaptation network to remove batch effects within the reference dataset. Additionally, it enhances the annotation of less frequent cell types by synthetic minority oversampling and ﬁts query data with the reference data using an autoencoder.

Study Duration

Not specified

Participants

1.7 million cells covering 230 human cell types

Evidence Level

Not specified

Key Findings

1
SELINA combines SMOTE, MADA, and an autoencoder to improve annotation accuracy.
2
SELINA pre-builds a reference atlas with 1.7 million cells covering 230 human cell types.
3
SELINA annotates cell types with high accuracy in various disease scenarios.

Research Summary

The rapid accumulation of single-cell RNA-seq data has provided rich resources to characterize various human cell populations. Here, we introduce SELINA (single-cell identity navigator), an integrative and automatic cell-type annotation framework based on a pre-curated reference atlas spanning various tissues. Notably, SELINA could accurately annotate cells within diverse disease contexts.

Practical Implications

Improved Cell-Type Annotation

SELINA provides a more accurate and robust method for cell-type annotation compared to existing tools, enabling better interpretation of scRNA-seq data.

Enhanced Disease Understanding

The ability of SELINA to accurately annotate cells in disease contexts can facilitate a deeper understanding of disease mechanisms and potential therapeutic targets.

Comprehensive Reference Atlas

The pre-curated reference atlas included in SELINA offers a valuable resource for researchers, providing a standardized and comprehensive dataset for cell-type identification.

Study Limitations

1
Insufﬁcient training cells and the limited number of datasets and cell types will lead to low annotation accuracy.
2
The reference still suffers from imbalances in tissue-speciﬁc datasets and cell-type abundance.
3
The deep learning model lacks biological insights such as identifying key factors for a cell type.

Your Feedback

Was this summary helpful?

Back to Bioinformatics