xCell2

xCell2 performs cell type enrichment analysis by generating cell type gene signatures from custom reference datasets and quantifying cell type proportions in bulk gene expression data.

Key Features:

• Custom Reference Training: xCell2Train generates cell type gene signatures using an improved methodology with automated handling of cell type dependencies and more robust signature generation from bulk or single-cell RNA-seq datasets.
• Cell Type Enrichment Analysis: xCell2Analysis conducts cell type enrichment analysis using trained custom references to estimate cell type proportions.
• Lineage Dependency Identification: xCell2GetLineage identifies dependencies between different cell types using ontology.
• Minimized Spillover Effects: The algorithm demonstrates superior performance in minimizing spillover effects between related cell types compared to other deconvolution methods.
• Pre-trained Signatures: The package includes a large collection of pre-trained cell type signatures for human and mouse research.

Scientific Applications:

• Tumor Microenvironment Characterization: xCell2-derived tumor microenvironment features can be fed into machine learning models such as LightGBM with nested cross-validation to predict immune checkpoint blockade response in cancer patients.
• Cross-species Deconvolution: The method supports cell type proportion estimation across diverse biological contexts using human and mouse reference datasets.
• Precision Medicine: Improved prediction of immune checkpoint blockade responses highlights potential applications for advancing precision medicine in cancer and other diseases.

Methodology:

xCell2 generates cell type gene signatures using an improved methodology including automated handling of cell type dependencies and more robust signature generation. The method was benchmarked against ten popular deconvolution methods using nine human and mouse reference sets and 26 validation datasets encompassing 1,749 samples and 67 cell types, and validated using the independent Deconvolution DREAM Challenge dataset.