Causal Relationship Analysis

bfLEAP®

AI-powered causal network discovery for drug discovery

Go beyond correlation. Uncover causal relationships. Discover meaningful disease drivers, biological signals, and patient populations hidden inside complex clinical and biological data.

~ 0 %

of drug candidates entering clinical development fail to achieve regulatory approval.¹

Most AI platforms in drug development identify correlations, inferring relationships and patterns in data. Most don’t build analytical tasks into a structured approach that runs a number of high-precision iterations required to produce a stability-validated result.

THE INPUT

What You Provide

Your team provides the data. No predefined hypotheses required.

Datasets may need to be harmonized by bfPREP™ depending on robustness of your existing data infrastructure.

THE WORK

What We Do

bfLEAP® replaces broad, single pass inference with structured, targeted comparisons, and stability-validated analysis.

Compare

Evaluate similarities and differences across patients, cohorts, genes, pathways, and outcomes.

Cluster

Identify biologically-distinct patient populations and hidden subgroups.

Prioritize

Separate meaningful signals from background noise and statistical artifacts.

Explain

Reveal the biological features driving differences between groups.

Validate

Confirm findings through repeated stability-tested analyses rather than one-time observations.

Not every pattern is meaningful. Our technology focuses on the findings that remain consistent when tested repeatedly. This targeted, structured analytical approach produces findings that are more reliable, and with better processing efficiency, than conventional bulk-analysis methods. We leverage AI-powered, proprietary algorithms:

Stability-Selection Ensemble Clustering

Runs hundreds of resampling iterations to confirm subgroups are statistically consistent.

Causal Driver Discovery

Separates disease-driving mechanisms from associated downstream consequences.

Scientific Interrogation

Scientists review biological features, stability metrics, and supporting evidence driving every signal.

Experts In The Loop

Scientists remain firmly in control of every outcome.

bfLEAP® surfaces findings you may have missed. Your team evaluates them.

This is not a black-box delivering recommendations. It is an analytical system designed for scientific review.

Subgroups explained

Every patient or network cluster is associated with defined biological features.

Signals reviewed

Findings are evaluated against known biology and scientific literature.

Reproducibility checked

Stability metrics are shared for review.

Results interrogated

Researchers can easily challenge, refine, and validate discoveries.

Decisions guided

Scientists determine which findings warrant further investment.

The Output

What You Receive

Not just patterns, but evidence-based causal network insights from biological and clinical data that provide you with meaningful subgroups and driver responder identification.

The outputs together answer the questions that matter for the success of your clinical trials or drug discovery programs.

The Outcome

What it Supports

Precision targeting, biomarker discovery, and clinical trial optimization without the unreliability of general purpose black-box models.

bfLEAP® has the ability to identify clinically-meaningful signals in extremely high-dimensional biological data, surfacing targets and biomarkers that conventional approaches miss, while remaining fully interpretable by the research team.

Proof Points

Case 1: Schizophrenia Biomarker Discovery (CATIE Trial)

~200 patients/drug, 30+ clinical attributes, 700,000+ genetic variables

Uncovered novel genetic variants tied to olanzapine response (including neuronal maturation genes) missed by standard statistical tools.

Case 2: Pancreatic Cancer Phase III Re-analysis (In collaboration with Eleison Pharmaceuticals & H. Lee Moffitt Cancer Center)

~10,000 pages of fragmented clinical trial PDF reports harmonized with bfPREP™, then analyzed with bfLEAP®

Uncovered 4 reproducible patient subgroups in a previously failed Phase III trial. Cluster A demonstrated a ~2.6× increase in mean survival on glufosfamide vs. best supportive care (149.3 vs. 58.2 days, p=0.065), aligned with the drug’s glucose-mediated mechanism. Findings confirmed by oncology experts and presented at ASCO GI 2026.

Feasibility Assessment

Request a feasibility assessment with our data or your own to explore what’s possible.

Reference 1: (Khairil, L., Benny, K. H. S., Jerry, J., Khatib, F. M., Che Ramli, M. D., & Kumar, S. (2026). AI in Drug Discovery: Clinical Failures, Regulatory Reality, and the Validation Crisis Behind the Hype. Pharmaceuticals, 19(6), 916. https://doi.org/10.3390/ph19060916)