Enabling cross-indication protein expression analysis using a curated pan-cancer dataset and a tailored workflow

Area of Science:

• Proteomics
• Cancer Biology
• Bioinformatics

Background:

• The National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium (CPTAC) generated comprehensive multi-omics data for over 1,000 tumors.
• Comparing protein expression across CPTAC cohorts is difficult due to missing data and varied expression patterns.

Purpose of the Study:

• To create a curated and normalized pan-cancer protein expression dataset from CPTAC data.
• To enable robust cross-cohort protein expression analysis for the cancer research community.

Main Methods:

• Developed a novel algorithm for selecting robustly expressed proteins within CPTAC cohorts.
• Applied a cohort hybrid imputation approach for protein abundance values.
• Utilized intensity-based absolute quantification and compared global vs. smooth quantile normalization.

Main Results:

• Global quantile normalization showed superior performance compared to smooth quantile normalization and no normalization.
• Higher rank correlation across cancer cohorts was observed between CPTAC and TCGA using global quantile normalization.
• The proposed workflow effectively addresses missing data and normalizes protein expression patterns.

Conclusions:

• Combining cohort hybrid imputation with global quantile normalization creates an effective normalized CPTAC pan-cancer protein dataset.
• This normalized dataset will facilitate the study of protein expression across diverse cancer types.
• The findings accelerate pan-cancer discovery research by enabling reliable proteomic data comparison.