Detecting Differentially Variable MicroRNAs via Model-Based Clustering.

Area of Science:

• Genomics
• Epigenetics
• Biostatistics

Background:

• Identifying differentially variable (DV) genomic probes is a novel approach for detecting genomic risk factors in complex diseases.
• The F-test is a standard statistical method for detecting differences in variance, successfully applied to DNA methylation data.
• MicroRNAs (miRNAs) are key epigenetic regulators, but DV miRNAs have not been previously identified.

Purpose of the Study:

• To propose a novel model-based clustering method to enhance the detection power of the probe-wise F-test for identifying DV miRNAs.
• To apply this method to discover novel DV miRNAs relevant to human diseases.

Main Methods:

• Developed a model-based clustering method incorporating specific covariance matrix structures.
• Imposed prior information on the relationship and independence of variances between cases and controls.
• Validated the method using simulation studies and analyzed two real-world datasets for human hepatocellular carcinoma (HCC).

Main Results:

• Simulation studies indicated the proposed method is promising for detecting DV probes.
• Identified seven DV-only miRNAs in HCC datasets: hsa-miR-1826, hsa-miR-191, hsa-miR-194-star, hsa-miR-222, hsa-miR-502-3p, hsa-miR-93, and hsa-miR-99b.
• Discovered hsa-miR-1826 as a potentially novel miRNA associated with HCC, not previously reported in the literature.

Conclusions:

• The proposed model-based clustering method effectively improves the detection of DV miRNAs.
• This novel approach has potential for identifying new epigenetic risk factors for complex diseases like HCC.
• The findings highlight specific miRNAs, including a newly implicated one, in the context of HCC pathogenesis.