Data aggregation at the level of molecular pathways improves stability of experimental transcriptomic and proteomic...

Area of Science:

• Biomedical research
• Genomics
• Bioinformatics

Background:

• High-throughput technologies enable massive gene expression analysis at RNA and protein levels.
• Expression data from different experiments often exhibit poor compatibility, even for identical biological samples.
• Cross- and intra-platform bias are significant challenges in analyzing gene expression data.

Purpose of the Study:

• To investigate methods for diminishing cross- and intra-platform bias in gene expression data.
• To develop and evaluate a mathematical model for aggregating gene expression data at the molecular pathway level.
• To compare the performance of different methods in data aggregation and retention of biological features.

Main Methods:

• Experimental and bioinformatic investigation of major gene expression platforms.
• Aggregation of gene expression data at the molecular pathway level.
• Development of a mathematical model for cumulative suppression of data variation.
• Comparison of five alternative data aggregation methods based on performance and feature retention.

Main Results:

• Aggregation of gene expression data at the molecular pathway level effectively diminishes cross- and intra-platform bias.
• A mathematical model was created to predict optimal parameters and pathway size for data aggregation.
• The OncoFinder bioinformatic method demonstrated superior performance in data aggregation and biological feature retention compared to four other methods.

Conclusions:

• Aggregating gene expression data by molecular pathways is a robust strategy to overcome platform-specific biases.
• The OncoFinder method offers optimal performance for cross-platform gene expression data analysis.
• This approach will be highly valuable for future large-scale, multi-platform biomedical data integration and analysis.