HieRFIT: a hierarchical cell type classification tool for projections from complex single-cell atlas datasets

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables detailed study of biological variation.
• Accurate cell classification is crucial for understanding cellular function and heterogeneity.
• Existing methods face challenges in precise cell type identification, especially across datasets.

Purpose of the Study:

• To introduce Hierarchical Random Forest for Information Transfer (HieRFIT), a novel computational tool for cell type classification.
• To enhance the accuracy and reliability of cell type projection using scRNA-seq data.
• To leverage hierarchical relationships between cell types for improved classification.

Main Methods:

• Development of HieRFIT, a tool based on hierarchical random forests.
• Integration of a priori information on cell type relationships via a hierarchical tree structure.
• Utilization of an ensemble approach combining multiple random forest models in a hierarchical decision tree.

Main Results:

• HieRFIT demonstrates improved classification accuracy, particularly for inter-dataset tasks.
• The hierarchical approach reduces incorrect cell type predictions.
• A scoring scheme adjusts probability distributions and resolves uncertainties, preventing misclassification.

Conclusions:

• HieRFIT offers a robust method for accurate cell type classification from scRNA-seq data.
• The tool's hierarchical approach effectively handles complex cell type relationships.
• HieRFIT provides a valuable resource for biological variation analysis and cell classification in bioinformatics.