FSQN

FSQN performs feature-specific quantile normalization to remove platform-based technical bias and enable integration of RNA-seq with DNA microarray data for cross-platform gene expression analysis and molecular subtype classification.

Key Features:

• Normalization Across Platforms: FSQN removes platform-based bias from RNA-seq data to allow consistent comparison across different gene expression profiling technologies.
• R package implementation: FSQN is implemented as an R package.
• Machine Learning Integration: The method employs machine learning classifiers trained on DNA microarray data to assign molecular subtypes to normalized RNA-seq datasets.
• High Accuracy Rates: FSQN achieved up to 98% accuracy for breast invasive carcinoma (BRCA) and 97% accuracy for colorectal cancer (CRC) in subtype classification.
• Versatility with Feature Scaling: The method maintains high performance regardless of feature scaling or the choice of machine learning algorithm.
• Sample Size Considerations: Maximum accuracy is achieved when RNA-seq datasets contain at least 25 samples.

Scientific Applications:

• Molecular subtype analysis in cancers: Integration of RNA-seq and DNA microarray data for subtype classification in breast invasive carcinoma (BRCA) and colorectal cancer (CRC).
• Cross-platform studies of disease biology: Normalization and classification across platforms to study molecular heterogeneity, disease pathogenesis, and therapeutic responses in cancers and autoimmune diseases.

Methodology:

FSQN applies feature-specific quantile normalization and trains machine learning classifiers on DNA microarray data to normalize RNA-seq datasets and assign microarray-derived molecular subtypes.