Juxtapose

Juxtapose compares gene co-expression networks (GCNs) across species, tissues, and experimental conditions to identify conserved and divergent network features.

Key Features:

• Implementation: Python-based implementation that leverages methodologies derived from natural language processing.
• Gene Embedding Strategy: Generates gene embeddings via random walks through GCNs combined with word-embedding techniques to represent network topology.
• Similarity Measures: Assesses topological similarity between networks using embedding-based measures such as cosine distance.
• Global Network Alignment: Performs global alignment of synthesized and real GCNs to identify conserved regions without relying on external biological information.
• Biological Evaluation: Validates comparisons using RNA-seq datasets such as GTEx and multi-species prefrontal cortex experiments, together with gene set enrichment analysis and literature-known gene relationships.
• Post-hoc Analysis: Supports downstream analyses incorporating biological parameters including gene orthology and conserved or variable pathways.

Scientific Applications:

• Evolutionary Studies: Comparative analysis of GCNs across species to investigate evolutionary conservation of gene regulatory relationships.
• Intra-species Transcriptomics: Comparison of co-expression networks across tissues or conditions within a species to identify conserved and condition-specific modules.

Methodology:

Generates gene embeddings from random walks on GCNs using word-embedding techniques, compares embeddings with cosine distance, performs global network alignment, and evaluates results with RNA-seq datasets (e.g., GTEx, multi-species prefrontal cortex) and gene set enrichment analysis.