VICTOR: Validation and inspection of cell type annotation through optimal regression

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity.
• Automated cell annotation is crucial for scRNA-seq data analysis.
• Reliability assessment of automated cell annotations, especially for rare cell types, remains a challenge.

Purpose of the Study:

• To introduce VICTOR (Validation and Inspection of Cell type annotation Through Optimal Regression).
• To develop a robust method for assessing the confidence of automated cell type annotations.
• To improve the reliability of single-cell data analysis.

Main Methods:

• Developed VICTOR, a method employing elastic-net regularized regression.
• Implemented optimal thresholds for gauging annotation confidence.
• Validated VICTOR across diverse single-cell datasets (within-platform, cross-platform, cross-study, cross-omics).

Main Results:

• VICTOR effectively identifies inaccurate cell type annotations.
• VICTOR demonstrates superior diagnostic ability compared to existing methods.
• The method shows robust performance across various experimental settings and data types.

Conclusions:

• VICTOR provides a reliable approach to assess cell annotation confidence.
• The tool enhances the accuracy and trustworthiness of single-cell data interpretation.
• VICTOR is a valuable addition to the single-cell analysis toolkit.