HieRFIT

HieRFIT classifies single-cell RNA sequencing (scRNA-seq) data using hierarchical random forest models to improve cell-type projection accuracy across datasets and species.

Key Features:

• Hierarchical Classification Approach: Uses a hierarchical decision-tree composed of multiple random forest models to leverage relationships among cell types for classification.
• A Priori Information Utilization: Requires a user-provided hierarchical tree representing class relationships and reference data as input to guide classification.
• Improved Inter-dataset Accuracy: Enhances performance for inter-dataset projection and cross-comparison of scRNA-seq data.
• Scoring Scheme for Uncertainty Resolution: Adjusts probability distributions for candidate class labels and resolves uncertainties, allowing assignment to internal hierarchy nodes when necessary.
• Cross-species and Intra-species Projection: Demonstrated applicability for both inter- and intra-species cell type cross-projections.

Scientific Applications:

• Cell-type classification: Classifies cells into functional and biologically meaningful categories using scRNA-seq data.
• Cross-study data integration: Projects and compares cell-type annotations across different datasets to support integrative single-cell analyses.
• Comparative and evolutionary studies: Enables projection of cell types across species to examine cellular heterogeneity and biological variation.

Methodology:

Constructs a hierarchical tree of cell-type relationships used with reference datasets to train an ensemble of random forest models organized hierarchically; applies a scoring scheme that adjusts probabilities and permits internal-node labeling to resolve classification uncertainty.