SPONGE

SPONGE constructs competing endogenous RNA (ceRNA) networks by modeling how multiple microRNAs (miRNAs) collectively regulate target transcripts to quantify miRNA-mediated cross-talk.

Key Features:

• multiple sensitivity correlation: Implements multiple sensitivity correlation, a quantitative measure that captures the joint contribution of multiple miRNAs to a ceRNA interaction.
• Probabilistic confounder adjustment: Employs a probabilistic model that adjusts for confounders when estimating ceRNA interactions.
• Analytical null distribution and P-value computation: Enables analytical estimation of null distributions to allow fast and accurate P-value computation.
• Scalability: Optimized for scalability to enable rapid inference of ceRNA networks in large transcriptomic datasets.
• Input data compatibility: Operates on paired miRNA and gene expression profiles, demonstrated on The Cancer Genome Atlas (TCGA) datasets.
• Detection of coding and non-coding ceRNAs: Detects both protein-coding and non-coding transcripts acting as ceRNAs and highlights candidate cancer biomarkers.

Scientific Applications:

• Genome-wide ceRNA network inference: Enables genome-wide inference of miRNA-mediated regulatory interactions from transcriptomic data.
• Biomarker discovery in cancer: Identifies known and novel ceRNAs, including protein-coding and non-coding transcripts, as candidate cancer biomarkers in TCGA datasets.
• Characterization of miRNA-mediated cross-talk: Quantifies miRNA-mediated cross-talk among transcripts by modeling multi-miRNA regulatory effects.

Methodology:

Computational components include the multiple sensitivity correlation metric and a probabilistic model that adjusts for confounders and enables analytical estimation of null distributions for fast P-value computation.