A probabilistic approach for automated discovery of perturbed genes using expression data from microarray or RNA-Seq.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Complex diseases involve alterations in multiple molecular and cellular components.
• Gene expression analysis reveals patient variability, yet disease progression suggests common cellular subprocess fates.
• Interconnected cellular subprocesses are key to understanding disease physiology.

Purpose of the Study:

• To develop an automated methodology for probing connected cellular processes in complex diseases.
• To identify disease-related genes (DRG) by assessing the probability of gene expression change (POC).
• To evaluate the influence of diseases on gene expression at network and pathway levels.

Main Methods:

• Combined biological networks, statistical models, and game theory for analysis.
• Utilized probability of change (POC) to quantify gene expression alterations between conditions.
• Applied machine learning to classify breast cancer subtypes using identified DRG.

Main Results:

• Identified DRG between breast cancer subtypes using RNA-Seq and microarray data.
• Developed a machine-learning algorithm for subtype discrimination, yielding a set of biomarkers.
• Achieved high agreement (100% for microarray, 80% for RNA-Seq) with PAM50 for disease subtype classification.

Conclusions:

• Presented an automated probabilistic approach for DRG and biomarker discovery.
• The method provides unbiased and reproducible results.
• Complements existing methods for analyzing complex diseases.