VCNet

VCNet constructs gene co-expression networks from high-dimensional RNA-seq data using vector-based statistical methods and a Frobenius-norm-based correlation test to detect exon-level gene-gene associations.

Key Features:

• Dimensional Problem Resolution: VCNet constructs GCNs reliably when the number of exons exceeds sample size, addressing the small-sample high-dimensional setting that challenges methods like SpliceNet and RNASeqNet.
• Statistical Hypothesis Testing: VCNet employs a hypothesis test based on the Frobenius norm of the correlation matrix for gene-gene pairs to identify network edges.
• Asymptotic Distribution: The method derives the asymptotic distribution of the test statistic under the null model to support inference.
• Performance Superiority: Simulation studies reported that VCNet outperforms SpliceNet and RNASeqNet in detecting meaningful edges within GCNs.
• Biological Relevance: Application to The Cancer Genome Atlas (TCGA) datasets produced networks with more biologically significant interactions compared to other methods.

Scientific Applications:

• RNA-seq Data Analysis: VCNet is applicable to RNA-seq analyses at the exon level for constructing gene co-expression networks from high-dimensional expression matrices.
• Biomedical Research: VCNet facilitates identification of biologically meaningful gene interactions for studies in oncology and genomics, as demonstrated on TCGA data.

Methodology:

VCNet leverages vector-based methods for high-dimensional RNA-seq data, computes correlation matrices, applies a Frobenius-norm-based statistical hypothesis test on gene-gene correlation matrices, and derives the test's asymptotic null distribution to detect network edges.