bfLEAP

bfLEAP® is an award-winning explainable artificial intelligence (AI) and machine learning platform developed at Johns Hopkins University Applied Physics Laboratory and exclusively licensed to BullFrog AI for use in drug development. Built as a graph analytic AI platform, bfLEAP® rapidly detects anomalies and uncovers hidden associations within disparate, multimodal, and incomplete data sets, leveraging both supervised and unsupervised machine learning to deliver unbiased, transparent, and explainable results.

Designed for researchers, clinicians, and drug development professionals, bfLEAP® focuses on biological and clinical data sets, transforming raw data into comprehensive data networks ready for in-depth analysis. The platform is engineered to accelerate the development of precision medicines by helping teams ethically unlock invaluable insights from academic, public, and proprietary program data.

Applications in Drug Development

• Drug discovery: By leveraging disease model data, bfLEAP® can identify new therapeutic targets, optimize drug combinations, and streamline the overall drug development process.
• Patient segmentation: The platform aids in segmenting patients, understanding dysregulated pathways, and tailoring treatments to specific populations.
• Drug repurposing: bfLEAP®'s advanced analytical capabilities can unearth potential expansion opportunities for existing drugs, enabling any molecule to realize its full therapeutic potential.
• Target discovery: By modeling multimodal data, bfLEAP® provides insights into associations between specific pathways, sets of genes, and disease conditions, moving beyond individual targets to understand complex biological system interplay.
• Clinical trial analysis: The platform supports optimization of inclusion/exclusion criteria, identification of patients most likely to respond to treatments, and efficient monitoring for efficacy and adverse events.

Core Capabilities

• Graph-based analysis: Unlike traditional linear models, bfLEAP®'s graph-based approach provides a holistic and accurate representation of the interconnected nature of genes, pathways, and diseases.
• Probability models: The platform excels at statistical analysis and model building, featuring a suite of state-of-the-art models including the proprietary Random Subspace Mixture Model (RSMM).
• Time series analysis: bfLEAP® includes functionalities for ingesting, enriching, and projecting time-series data into graphs, enabling tracking of disease progression, trend understanding, and predictive analyses based on historical data.
• Layered analysis: Complex processes are broken down into interpretable layers, ensuring researchers always understand how the AI arrived at its conclusions.
• Node neighborhood exploration: Biological and clinical data converted into graph format offers a visual method to interpret results and clarify relationships.
• Feature clustering: Users can group similar data points — whether features or samples — to identify patterns or anomalies.
• Ethical outputs: Data context is preserved throughout analysis, ensuring results accurately represent real-world significance.

Explainable AI: Transparency at the Core

• Unlike traditional AI systems such as neural networks that operate as black boxes, bfLEAP® prioritizes transparency so that every analytical output can be interpreted and trusted.
• Explainable results support ethical research and development practices.
• Transparent outputs enable informed decision-making based on AI recommendations.
• Explainability builds trust and confidence in AI-driven insights, which is essential in healthcare and clinical research contexts.

RSMM Anomaly Detection

• Benchmark excellence: In rigorous tests using 12 open-source data sets, the RSMM algorithm consistently outperformed leading models in the field, including generative AI Variational Autoencoders (VAEs).
• Contextual interpretation: Beyond identifying anomalies, RSMM provides contextual interpretation of their significance — particularly important for biologists and clinicians.
• Quality control: Identifies potential errors, outliers, or inconsistencies to ensure data integrity.
• Safety monitoring: Highlights safety deviations of greatest relevance in clinical trials and marketed products.
• Patient subgroup identification: Uncovers unique patient subgroups to facilitate tailored treatments and therapies.
• Early efficacy detection: Highlights signs of treatment success at the earliest stages, enabling more efficient trial evaluations.
• Novel pattern discovery: Unearths new patterns that may lead to innovative hypotheses and future breakthroughs.

bfLEAP® is purpose-built for the drug development industry and is capable of working with academic data, public data sets, and proprietary program data. Its foundation in explainable AI, combined with pioneering anomaly detection and graph-based analytics, makes it a comprehensive platform for professionals seeking clarity, efficiency, and precision throughout the drug discovery and development lifecycle.