CLIC

CLIC (CLustering by Inferred Co-expression) infers co-expression modules to predict novel pathway members and identify the RNA expression datasets in which those pathways are active.

Key Features:

• Co-Expression Analysis: CLIC uses a Bayesian partition model to partition input gene sets into coherent co-expressed modules (CEMs) by analyzing expression patterns across multiple datasets.
• Posterior Probability Assignment: For each CEM, CLIC assigns a posterior probability per dataset quantifying support for the module's existence.
• Module Expansion: CLIC expands each CEM by scanning the transcriptome for additional genes with co-expression using an integrated log-likelihood ratio (LLR) score weighted by dataset relevance.
• Contextual Learning: CLIC learns and identifies the specific conditions or datasets in which each CEM is active.
• Extensive Dataset Utilization: Implementation integrates a compendium of 1774 mouse microarray datasets (28,628 microarrays) and 1887 human microarray datasets (45,158 microarrays).

Scientific Applications:

• Pathway component discovery: CLIC predicts new components of biological pathways by identifying genes co-expressed with well-characterized pathway members.
• Hypothesis generation and validation: CLIC's predictions can generate experimental hypotheses, exemplified by the prediction linking C7orf55 (FMC1) to the mitochondrial ATP synthase complex that was subsequently validated.

Methodology:

Accepts a predefined set of pathway genes as input; applies a Bayesian partition model to identify co-expressed modules (CEMs); assigns posterior probabilities to datasets supporting each module; expands modules by scanning the transcriptome and scoring additional genes with an integrated LLR weighted by dataset relevance; and learns the conditions or datasets in which modules are active.