Regulus infers signed regulatory relations from few samples' information using discretization and likelihood...

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Transcriptional regulation governs gene expression via transcription factors (TF) binding to DNA.
• Current methods for inferring regulatory circuits require extensive samples and struggle to predict activation/inhibition.
• Limited sample availability in human studies necessitates more efficient inference methods.

Purpose of the Study:

• To develop a method for inferring transcriptional regulatory circuits from fewer samples.
• To integrate diverse biological data for robust TF-gene relation prediction.
• To accurately qualify inferred relations as gene activation or inhibition.

Main Methods:

• Regulus integrates TF binding, gene expression, and regulatory region accessibility data.
• Data complexity is reduced by aggregating into patterns and storing in a Resource Description Framework (RDF) endpoint.
• Biology-based likelihood constraints and SPARQL queries filter TF-region-gene relations and qualify them as activation or inhibition.

Main Results:

• Regulus computes TF-gene relations, providing signed predictions (activation/inhibition).
• The method yields results consistent with public databases.
• Application to biological data identified known and novel transcriptional regulators.

Conclusions:

• Regulus enables context-specific transcriptional circuit inference, particularly in low-sample human settings.
• The method's pattern discretization and likelihood reasoning identify robust regulatory relations.
• Regulus enhances the discovery of gene regulatory mechanisms.