EdgeScaping

EdgeScaping: Gene Co-Expression Network Construction and Non-Linear Analysis

EdgeScaping constructs and analyzes gene co-expression networks (GCNs) from Gene Expression Matrices (GEMs) using a top-down strategy that retains all pairwise gene expression data without edge filtering to enable scalable detection of non-linear gene relationships.

Key Features:

• Image-Based Data Conversion: Transforms pairwise gene expression data into an image-based representation for feature compression and scalable analysis of large GEMs.
• Deep Learning Integration: Applies image-based deep learning algorithms to detect complex non-linear gene expression patterns beyond standard correlation tests.
• Retention of Non-Conventional Edges: Preserves all network edges, including those typically discarded as noisy, to capture unconventional co-expression relationships.
• Efficient Clustering Algorithm: Clusters gene expression relationships in the embedded feature space derived from image-based conversion while preserving gene expression intensity context.

Scientific Applications:

• Human Tumor GEM Analysis: Identifies conventional and non-conventional interdependent non-linear gene behaviors in human tumor Gene Expression Matrices, including brain-specific tumor sub-types.

Methodology:

EdgeScaping converts pairwise gene expression data from Gene Expression Matrices into an image-based format to compress features and enable scalability. Deep learning algorithms analyze the transformed data to cluster and interrogate the full space of gene expression relationships while retaining expression intensity context for biological interpretation.