ADAPTS: Automated deconvolution augmentation of profiles for tissue specific cells.

Area of Science:

• Computational Biology
• Bioinformatics
• Immunology

Background:

• Immune cell infiltration and the tumor microenvironment are critical for patient outcomes.
• Gene expression deconvolution helps identify patient risk and treatment strategies.
• Signature matrices derived from single-cell data are key to deconvolution methods.

Purpose of the Study:

• To introduce ADAPTS, a framework for cell type deconvolution.
• To provide modular tools for customizing and creating signature matrices.
• To offer a common interface for popular deconvolution algorithms.

Main Methods:

• ADAPTS implements tools for customizing signature matrices (adding cell types, de novo creation from single-cell RNAseq).
• It integrates popular deconvolution algorithms using signature matrices.
• A novel hierarchical deconvolution method is included for clustering and re-splitting cell types.

Main Results:

• ADAPTS improves the reconstruction of cell types from single-cell RNAseq data.
• Demonstrated enhanced predictive analysis in a blind setting.
• The package provides a flexible framework for various deconvolution needs.

Conclusions:

• ADAPTS offers a comprehensive solution for cell type deconvolution in complex biological samples.
• The package enhances the utility of signature matrices and deconvolution algorithms.
• ADAPTS is available in R, facilitating broader research in tumor microenvironment analysis.