bAIcis

bAIcis learns and simulates Bayesian network (BN) structures from large-scale observational datasets to model dependencies among discrete, continuous, and mixed variables for applications such as genomics and multi-omics research.

Key Features:

• Comprehensive Performance: Evaluated against existing open-source BN learners and reported superior structure recovery with a true positive rate of 0.9 and precision of 0.8.
• Versatility Across Data Types: Supports discrete, continuous, and mixed variables for heterogeneous data analyses.
• Scalability and Efficiency: Parallelized on distributed systems to manage datasets with thousands of features while maintaining accuracy.
• Applicability to Large Feature Spaces: Capable of handling feature spaces that can exceed hundreds of thousands, enabling analyses in genomics and multi-omics research.

Scientific Applications:

• Genomics: Modeling relationships among numerous genes to infer genetic networks.
• Multi-Omics Research: Integrating omic layers (e.g., genomics, proteomics) to uncover complex biological interactions and pathways.
• Healthcare Analytics: Recovering precise network structures from patient data to inform predictive models in personalized medicine.

Methodology:

Learns and simulates Bayesian networks, is parallelized on distributed systems, and was evaluated using synthetic datasets containing discrete, continuous, and mixed data across small and large feature spaces and benchmarked against open-source BN learners reporting a true positive rate of 0.9 and precision of 0.8.