sem1R

sem1R induces semantically coherent patterns and interpretable rules from 2-dimensional binary omics data using an ontology-based refinement operator to reveal meaningful biological patterns.

Key Features:

• Data input: Operates on 2-dimensional binary omics data.
• Ontology-based refinement operator: Leverages prior knowledge encoded in ontologies to generate accurate and interpretable rules.
• CN2 inspiration: Implements a refinement operator inspired by the CN2 rule learner.
• Reduction procedures: Applies Redundant Generalization and Redundant Non-potential procedures to prune the rule space.
• Rule induction and biclustering: Induces interpretable rules and discovers semantically coherent biclusters.
• Class comparison: Captures semantic differences between a target class of positive examples and a non-target class of negative examples.
• Efficiency: Prunes the rule space to accelerate pattern induction relative to traditional refinement operators.

Scientific Applications:

• Semantic pattern discovery in omics: Identifies non-trivial, meaningful semantic patterns within omics datasets.
• Complement to Gene Set Enrichment Analysis (GSEA): Enables induction of complex patterns beyond analyses based on sorted gene lists and ontological annotations.
• Bicluster discovery: Reveals semantically coherent biclusters to support nuanced interpretation of biological systems.
• Validation on gene expression data: Demonstrated on the Dresden Ovary Dataset, DISC, and m2816 gene expression datasets.

Methodology:

Induces rules from 2-dimensional binary omics data using an ontology-based refinement operator inspired by CN2, applies Redundant Generalization and Redundant Non-potential reduction procedures to prune the rule space, and captures semantic differences between positive and negative classes.