Regulus

Regulus infers transcriptional regulatory networks by integrating transcription factor (TF) binding sites, gene expression levels, and regulatory region accessibility/activities using Semantic Web technologies (RDF and SPARQL) and biology-based logical constraints to predict signed (activation or inhibition) TF–gene interactions, including in settings with limited sample sizes.

Key Features:

• Integration of Multiple Data Types: Integrates TF binding sites, gene expression profiles, regulatory region accessibility/activities, and genomic locations to associate TFs with potential target genes.
• Data Aggregation and Pattern Recognition: Aggregates gene expression and regulatory region activities into discrete patterns to reduce data complexity and facilitate comparison across samples.
• Semantic Web Technologies: Uses RDF endpoints and SPARQL queries to retrieve candidate TF–gene relations involving active regulatory regions.
• Biological Logic Constraints: Applies biology-based logical constraints to ensure global consistency and to assign interaction signs as activation or inhibition.
• Performance and Validation: Demonstrates performance against existing network inference methods by providing signed relations that align with public databases and by identifying known and putative novel regulators.

Scientific Applications:

• Network inference from limited samples: Inferring transcriptional regulatory networks when sample sizes are limited and cell populations are closely related.
• Signed interaction assignment: Distinguishing activation versus inhibition relationships between TFs and target genes.
• Regulator identification and validation: Identifying established regulators and proposing candidate novel regulators for follow-up with comparison to public databases.

Methodology:

Combine TF binding data, gene expression profiles, and regulatory region activities with genomic locations; aggregate expressions and region activities into patterns; query RDF endpoints with SPARQL to extract candidate TF–gene relations involving active regions; apply biology-based logical filtering to qualify and sign inferred interactions.