Decomprolute is a benchmarking platform designed for multiomics-based tumor deconvolution.

Area of Science:

• Computational Biology and Bioinformatics
• Proteogenomics
• Cancer Research

Background:

• Tumor deconvolution identifies cell types in solid tumors, but current methods primarily focus on gene expression (e.g., RNA sequencing) rather than protein levels.
• Protein levels offer a more accurate representation of rare cell types compared to gene expression data.
• A need exists for standardized tools to evaluate deconvolution algorithms across diverse, multiomic datasets.

Purpose of the Study:

• To introduce Decomprolute, a Common Workflow Language (CWL) framework for comparing tumor deconvolution algorithms.
• To facilitate the development, use, and reproducibility of multiomic deconvolution algorithms.
• To enable benchmarking of algorithms using large-scale proteogenomic datasets.

Main Methods:

• Developed Decomprolute as an open-source, containerized CWL framework.
• Integrated large-scale multiomic datasets from the Clinical Proteomic Tumor Analysis Consortium (CPTAC).
• Included matched mRNA expression and proteomic data from thousands of tumors across multiple cancer types.

Main Results:

• Established a fully open-source, containerized platform for benchmarking proteogenomic tumor deconvolution algorithms.
• Enabled comparison of deconvolution algorithms across multiomic datasets, leveraging CPTAC data.
• Facilitated reproducible analysis of tumor cellularity using both gene expression and protein abundance.

Conclusions:

• Decomprolute provides a robust and reproducible platform for advancing multiomic tumor deconvolution.
• The framework supports the development and validation of algorithms that utilize protein-level data for improved cell type identification.
• This resource promotes standardized benchmarking in cancer proteogenomics research.