GANomics: bridging legacy and modern transcriptomic platforms for clinical applications

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Transcriptomic technologies have evolved, necessitating methods to integrate legacy microarray data with modern RNA-sequencing (RNA-seq) data.
• Leveraging historical transcriptomic datasets is crucial for comprehensive analysis and biomarker discovery.
• Existing methods may struggle with accurate data translation between different transcriptomic platforms.

Purpose of the Study:

• To develop a robust framework, GANomics, for bidirectional data translation between microarray and RNA-seq platforms.
• To enable the effective integration of historical and contemporary transcriptomic data while preserving biological consistency.
• To facilitate the reuse of biomarkers and expand transcriptomic resources for clinical applications.

Main Methods:

• Utilized a generative adversarial network (GAN) framework named GANomics.
• Integrated paired and unpaired samples using a pair-aware feedback loss mechanism.
• Enforced one-to-one transcript mappings and preserved global gene expression distributions during data translation.

Main Results:

• Achieved high per-sample correlations (>0.96) between real and synthetic data with as few as ten paired profiles.
• Accurately recapitulated differential gene expression and maintained pathway-level rankings with fifty paired profiles.
• Enabled a cross-platform classifier to transfer with performance comparable to real data, demonstrating robustness across multiple datasets.

Conclusions:

• GANomics effectively bridges legacy and contemporary transcriptomic data, retaining biological consistency.
• The framework facilitates scalable data integration for biomarker reuse and enhances transcriptomic resource expansion in clinical settings.
• GANomics offers a powerful solution for leveraging historical transcriptomic data alongside modern datasets.