HRM

HRM predicts gene dysregulation by integrating high-throughput RNA sequencing (RNASeq) data with prior regulatory network knowledge via the Host Response Model to infer transcriptional responses and the directionality and magnitude of gene expression changes to single and combined biochemical inducers.

Key Features:

• Machine Learning Integration: Employs a machine learning framework (Host Response Model, HRM) that integrates RNASeq data with known regulatory networks to infer transcriptional responses.
• Predictive Accuracy: Accurately predicts the directionality of gene dysregulation in response to combinations of inducers with over 90% accuracy.
• Enhanced Predictive Performance: Incorporation of prior regulatory network knowledge increases the coefficient of determination (R²) from 0.3 to 0.65, effectively doubling predictive capability.
• Validation Across Species: Validated in Escherichia coli and Bacillus subtilis, demonstrating applicability across different bacterial model organisms.
• Enrichment Analysis Capability: Enables enrichment analyses of predicted differential expression and classifies pathway regulation with over 95% accuracy.
• Experimental Efficiency: Supports in silico testing of responses to reduce the need for extensive RNASeq experiments.

Scientific Applications:

• Synthetic Biology: Predicts cellular transcriptional responses to synthetic constructs and modifications to inform design strategies.
• Systems Biology: Aids interpretation of complex regulatory networks by predicting how combinations of perturbations affect gene expression.
• Drug Discovery and Development: Models transcriptional responses to biochemical agents to assist identification of potential drug targets and assessment of off-target effects.

Methodology:

Train a machine learning model on RNASeq data from single-inducer experiments, integrate prior knowledge of regulatory networks (Host Response Model), predict responses to combinations of inducers, and perform enrichment analyses on predicted differential expression.