PXN Unlocks the Power of Public Gene Expression Data Through Cross-Technology Integration

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Public gene expression repositories are valuable but limited by incompatible datasets from diverse technologies.
• Differences in measurement scales and signal distributions create systematic discrepancies, hindering large-scale integrative analysis.
• Inconsistencies across platforms and labs impede studies requiring high statistical power and reproducibility.

Purpose of the Study:

• To introduce PXN, a probabilistic machine learning framework for a unified representation of biological signals across multiple gene expression technologies.
• To enable seamless data translation between platforms, preserving biological variation while removing technology-specific biases.
• To enhance the integration of heterogeneous gene expression datasets for improved biological discovery.

Main Methods:

• Developed a probabilistic machine learning framework (PXN) to capture a unified biological signal representation.
• Trained PXN to translate gene expression data across multiple platforms.
• Benchmarked PXN against existing normalization methods for cross-platform accuracy and differential expression analysis power.

Main Results:

• PXN consistently outperforms existing normalization methods in cross-platform accuracy.
• PXN substantially enhances the power of differential expression analysis.
• PXN successfully bridges the technological divide between microarray and RNA-seq data.

Conclusions:

• PXN provides a scalable solution for integrating legacy microarray data with modern RNA-seq studies.
• The framework enables direct comparison and integration of heterogeneous gene expression datasets.
• PXN unlocks the full potential of public repositories for future biological discovery and therapeutic innovation.