RCRnorm: An integrated system of random-coefficient hierarchical regression models for normalizing NanoString...

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Formalin-fixed paraffin-embedded (FFPE) samples are valuable for biomarker discovery and clinical studies.
• Traditional gene expression profiling methods perform poorly on damaged RNA from FFPE samples.
• Existing normalization methods for the NanoString nCounter platform are not optimized for FFPE samples.

Purpose of the Study:

• To develop an improved normalization method for gene expression profiling of FFPE samples.
• To address limitations of current normalization techniques for NanoString nCounter data.
• To facilitate the use of FFPE samples for biomarker discovery and clinical applications.

Main Methods:

• Developed a Bayesian approach using random-coefficient hierarchical regression models (RCRnorm).
• Integrated experimental design information to capture FFPE sample characteristics.
• Simultaneously removed biases from multiple sources without relying on housekeeping gene assumptions.

Main Results:

• RCRnorm effectively normalizes gene expression across FFPE samples.
• The method demonstrates superior performance compared to existing techniques in simulations and applications.
• RCRnorm is also applicable to freshly frozen samples.

Conclusions:

• RCRnorm provides a robust and interpretable normalization strategy for FFPE samples on the NanoString platform.
• The method enhances the reliability of gene expression profiling from FFPE tissues.
• This advancement supports broader utilization of FFPE samples in research and diagnostics.