ModEnt

ModEnt reconstructs gene regulatory networks from high-throughput gene-expression profiles by formulating error reduction as a minimum-entropy problem to avoid discretization and enable continuous representation of expression levels.

Key Features:

• Minimum-entropy formulation: Formulates error reduction in gene-expression data as a minimum-entropy problem to avoid discretization and represent expression continuously.
• Probabilistic formulation: Adopts a probabilistic formulation of logical models to resolve inconsistencies in real-valued measurements.
• Heuristic optimization: Uses a heuristic algorithm to minimize errors while maintaining model parsimony, balancing complexity and fidelity to data.
• Computational complexity acknowledgment: Recognizes that reconstructing an error-minimized logical GRN is NP-complete and addresses this via the probabilistic and heuristic approach.
• Input data: Operates on high-throughput gene-expression profiles without conversion to discrete on/off states.
• Biological insight: Produces reconstructed logical GRNs that reveal shared regulatory logic among co-regulated genes and improve consistency with biological knowledge.
• Evaluation dataset: Validated on mouse embryonic stem cell data, demonstrating improved network models compared with raw experimental measurements.

Scientific Applications:

• GRN reconstruction: Reconstruction of logical gene regulatory networks from continuous expression data.
• Regulatory logic analysis: Investigation of shared regulatory logic among co-regulated genes.
• Biological process study: Analysis of regulatory programs relevant to differentiation, metabolism, and cell cycle regulation.
• Data interpretation: Improving interpretation of high-throughput gene-expression datasets by reducing measurement inconsistencies.

Methodology:

Formulates error reduction as a minimum-entropy problem; employs a probabilistic formulation of logical models; applies a heuristic algorithm to minimize errors subject to model parsimony; notes that error-minimized logical GRN reconstruction is NP-complete.