xCell 2.0: robust algorithm for cell type proportion estimation predicts response to immune checkpoint blockade.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Accurate cell type proportion estimation from bulk gene expression is crucial for understanding tissue heterogeneity in diseases.
• Existing methods face challenges with diverse reference datasets and complex cellular dependencies.

Purpose of the Study:

• Introduce xCell 2.0, an improved algorithm for cell type deconvolution.
• Enhance the accuracy and robustness of cell type signature generation and estimation.

Main Methods:

• Developed xCell 2.0 with a novel training function adaptable to any reference dataset.
• Implemented automated handling of cell type dependencies for more robust signature generation.
• Benchmarked xCell 2.0 against eleven deconvolution methods using extensive human and mouse datasets.

Main Results:

• xCell 2.0 demonstrated superior accuracy and consistency compared to eleven popular deconvolution methods across diverse reference sets and biological contexts.
• The algorithm showed improved performance in minimizing spillover effects between related cell types.
• xCell 2.0-derived tumor microenvironment features significantly enhanced pan-cancer immune cell checkpoint blockage response prediction accuracy.

Conclusions:

• xCell 2.0 is a versatile and robust tool for cell type deconvolution, maintaining high performance across various reference types and biological contexts.
• The tool is accessible via a web application and a Bioconductor package.
• Pre-trained cell type signatures for human and mouse research are readily available.