A Bayesian algorithm for detecting differentially expressed proteins and its application in breast cancer research.

Area of Science:

• Biochemistry
• Bioinformatics
• Oncology

Background:

• Mass spectrometry (MS) based proteomic data analysis is challenging due to noise and missing values.
• Missing values in MS data arise from proteins with abundances below detection limits.
• Accurate analysis is crucial for understanding complex biological systems and diseases like cancer.

Purpose of the Study:

• To develop and validate a novel Bayesian algorithm for analyzing MS-based proteomic data with missing values.
• To identify differentially expressed proteins in breast cancer (BC) subtypes.
• To discover potential diagnostic or prognostic biomarkers for BC patient survival.

Main Methods:

• Developed a Bayesian algorithm that incorporates missing data as informative.
• Validated the algorithm using simulated datasets and compared its performance against existing methods.
• Applied the algorithm to analyze proteomic data from a cohort of breast cancer patients.

Main Results:

• The Bayesian algorithm consistently outperformed other methods in accuracy on simulated data.
• Significant differences in proteomic landscapes were observed between triple-negative and Luminal A breast cancer subtypes.
• The majority of observed differences were linked to the transcriptional activity of seven key transcription factors.
• Two novel proteins significantly correlated with breast cancer patient survival were identified.

Conclusions:

• The developed Bayesian algorithm offers a robust approach for handling missing data in MS-based proteomics.
• Proteomic differences between aggressive and non-aggressive breast cancer subtypes are substantial and potentially driven by specific transcription factors.
• The identified proteins hold promise as potential diagnostic or prognostic biomarkers for breast cancer.