ImmClassifier

ImmClassifier classifies immune cell types from single-cell RNA sequencing (scRNA-Seq) transcriptional profiles using a knowledge-based, lineage-driven approach to enable fine-grained annotation of immune populations.

Key Features:

• Knowledge-Based Approach: Utilizes a hierarchical ontology of immune cell types grounded in the biology of immune cell differentiation and integrates heterogeneous reference datasets.
• Lineage-Driven Classification: Incorporates lineage information to classify cells based on developmental pathways.
• Integration of Machine Learning Models: Employs a cascading architecture that combines machine learning and deep learning modules to improve classification accuracy and robustness.
• High Prediction Accuracy: Demonstrates improved recall (+20%) and precision (+14%) and distinguishes fine-grained types such as CD8+ effector memory T cells while maintaining comparable results for broader categories like T cells versus macrophages.

Scientific Applications:

• Fine-Grained Immune Profiling: Enables detailed characterization of diverse immune cell populations in scRNA-Seq data.
• Tumor and Inflamed Tissue Analysis: Supports profiling of complex tissue environments such as tumors and inflamed tissues.
• Simultaneous Immune and Non-Immune Cell Profiling: Profiles both immune and non-immune cells (e.g., cancer cells) within the same sample.
• Alternative to Cytometry: Addresses limitations of traditional flow and mass cytometry for resolving cellular heterogeneity in scRNA-Seq experiments.

Methodology:

Leverages heterogeneous reference datasets, a hierarchical ontology-based classification grounded in immune differentiation, and a cascading architecture that combines machine learning and deep learning modules to classify single-cell RNA-Seq transcriptional profiles.