SVM-RNE

SVM-RNE applies support vector machines with recursive network elimination to classify gene expression samples and identify biomarker gene modules by integrating gene interaction networks.

Key Features:

• Gene Network Integration: Utilizes large gene interaction network databases and maps expression data onto the network.
• Initial Gene Selection (t-test): Begins with selection of 1,000 genes from a training set using the t-test.
• Network Mapping Filter: Filters the selected genes to retain only those that map onto a gene network database.
• Clustering with GXNA: Applies the Gene eXpression Network Analysis tool (GXNA) to form clusters of highly connected genes within the network.
• Linear SVM Classification and Cluster Weighting: Uses Linear SVM to classify samples based on network-derived clusters and assigns weights to each cluster reflecting its importance.
• Recursive Network Elimination: Iteratively removes the least informative clusters while retaining others for subsequent classification steps.

Scientific Applications:

• Gene expression classification: Classifying microarray gene expression datasets using network-informed features.
• Biomarker identification: Identifying biomarker gene modules via network-based recursive feature elimination.
• High-dimensional transcriptomic analysis: Reducing dimensionality and increasing biological interpretability of classifiers for high-dimensional expression data.

Methodology:

Select 1,000 genes from a training set using the t-test; filter to genes that map onto a gene network database; apply GXNA to form clusters of highly connected genes; use Linear SVM to classify samples based on these clusters and assign weights to clusters; iteratively remove the least informative clusters and repeat classification until optimal classification is achieved.