NormiRazor

NormiRazor performs GPU-accelerated selection and assessment of reference miRNAs for normalization of miRNA expression data to reduce non-biological variability in qPCR and circulating miRNA studies.

Key Features:

• GPU acceleration: Implements parallel computation on a graphics processing unit using the CUDA platform.
• Parallelized algorithms: Provides parallel implementations of geNorm, NormFinder, and BestKeeper for reference evaluation.
• Multi-miRNA combination support: Evaluates averaged expressions of multi-miRNA combinations as candidate reference references to improve stability over single genes.
• Automatic reference selection: Selects reference miRNAs automatically based on the implemented normalization algorithms.
• Benchmarking on public datasets: Validated on publicly available miRNA expression datasets including subsets of GSE68314 with reported execution time reductions of 18.7 ±0.6× for geNorm, 104.7 ±4.2× for BestKeeper, and 76.5 ±2.2× for NormFinder versus prior Python implementations.
• Applicability to circulating miRNAs and multi-gene assays: Targets normalization challenges specific to circulating miRNAs and multi-gene expression assays.

Scientific Applications:

• qPCR miRNA normalization: Enables robust normalization of qPCR-based miRNA expression datasets to reduce technical variability.
• Circulating miRNA studies: Supports selection of stable reference miRNAs for circulating biomarker analyses.
• Reference gene evaluation: Facilitates systematic evaluation of single and multi-miRNA reference combinations to improve reproducibility of transcriptomic studies.
• Large-scale normalization analyses: Accelerates evaluation of many reference combinations in large datasets to enable extensive stability testing.

Methodology:

geNorm, NormFinder, and BestKeeper were implemented in parallel on CUDA-enabled GPUs; averaged expressions of multi-miRNA combinations were evaluated as candidate references and performance was benchmarked on subsets of GSE68314 with comparisons to prior Python implementations.