scReClassify: post hoc cell type classification of single-cell rNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables transcriptome profiling at the individual cell level.
• Accurate cell type identification is crucial for developmental biology, cell reprogramming, and cancer research.
• Current methods combining biological knowledge and computational techniques are prone to mislabelling due to incomplete knowledge and subjectivity.

Purpose of the Study:

• To develop a semi-supervised learning framework, scReClassify, for post hoc cell type identification in scRNA-seq data.
• To improve the accuracy of cell type annotations by reclassifying potentially mislabelled cells.
• To provide a robust tool for fine-tuning existing cell type classifications.

Main Methods:

• scReClassify employs Principal Component Analysis (PCA) for dimension reduction.
• A semi-supervised learning algorithm is applied to learn from initial annotations and reclassify cells.
• The framework was validated using both simulated and real-world scRNA-seq datasets from diverse biological systems.

Main Results:

• scReClassify accurately identifies and reclassifies mislabelled cells to their correct cell types.
• The method demonstrates effectiveness across various tissues and biological contexts.
• The tool successfully refines cell type annotations derived from any classification procedure.

Conclusions:

• scReClassify serves as a valuable post hoc tool for enhancing cell type annotations in scRNA-seq datasets.
• The R package is freely available, promoting its adoption in the research community.
• This approach helps overcome limitations in current cell type identification methods.