CellO

CellO classifies cell types from human RNA-seq data using the hierarchical Cell Ontology to provide standardized annotations for single-cell RNA-sequencing analyses.

Key Features:

• Ontology-Based Classification: Uses the Cell Ontology hierarchical framework to assign cell-type labels that respect biological relationships among human cell types.
• Machine Learning Models: Employs machine learning models trained on nearly all available human primary cell and bulk RNA-seq data from the Sequence Read Archive (SRA).
• Pre-Trained Models: Provides models pre-trained on healthy, untreated primary samples for application to single-cell RNA-seq datasets.
• Competitive Performance: Reports competitive or superior classification accuracy relative to existing state-of-the-art cell type annotation methods.
• Interpretable Linear Models: Uses linear models whose coefficients enable interpretation of cell-type-specific gene expression signatures.

Scientific Applications:

• Cell type annotation for single-cell RNA-seq: Produces standardized cell-type labels to characterize cellular composition and heterogeneity in single-cell datasets.
• Discovery of novel or uncharacterized cell populations: Facilitates identification of potentially novel cell types via ontology-aware classification.
• Comparative transcriptomics and systems biology: Enables downstream analyses that compare cell-type-specific expression across tissues or conditions in genomics and transcriptomics studies.

Methodology:

Implemented as a Python package that employs machine learning (linear) models trained on extensive human primary cell and bulk RNA-seq data from the SRA, with provided pre-trained models derived from healthy, untreated primary samples.