Literature is vast. Connections are invisible. We illuminate repurposing pathways.
Bane mines peer-reviewed medical publications to construct traceable, multi-hop knowledge graphs. Extracting direct mechanism-of-action chains from published papers—preserving opposing trials without generative hallucinations.
Billions spent discovering new molecules. Decades of clinical signals left unconnected.
Modern biomedical research is siloed across millions of isolated abstracts. Significant computational repurposing signals are already documented in published trials, waiting for structural synthesis.
Unstructured text traps 85% of biological linkages.
Standard keyword queries on PubMed return long bibliographies without resolving causal pharmacological links. Researchers are forced to manually correlate drug affinities with distant pathway mutations.
Generic LLMs collapse nuance and suppress negative trials.
Generative chat tools summarize research by smoothing over contradictions. In translational medicine, limiting trials and negative toxicity markers are as vital as positive efficacy claims.
Discovering repurposing opportunities through Swanson's ABC Model.
Bane mines literature for approved drugs and their biological targets. By establishing transitive graph linkages (A -> B -> C), we surface high-impact secondary repurposing opportunities without speculative guesswork.
Swanson's A → B → C Model: If Drug A modulates Target B, and Target B is a causal driver in Disease C, an indirect repurposing signal (A → C) is inferred and scored across literature citations.
The Researcher's Mobile Workbench
Explore the Bane client application interface. Navigate through the screens below using the arrow controls.
.png)
Full-stack computational intelligence, from abstract to candidate dossier.
Europe PMC REST Ingestion
Automated retrieval with query expansion, keyword relevancy filters, and deduplication across millions of peer-reviewed articles.
Dual Transformer Models
Fine-tuned Named Entity Recognition (NER) paired with token-level relation extraction classifying directional biological verbs.
NetworkX Graph Topology
Constructs interactive multi-hop knowledge graphs exported to PyVis HTML for webview inspection and topological analysis.
Scored Signals Workbench
Ranks candidate repurposing opportunities via composite confidence scoring, separating statistical priority from clinical proof.
Active Literature Surveillance
Real-time alert feeds that monitor emerging publications for target mutations, clinical trials, and newly extracted candidate signals.
Cross-Platform Flutter Client
High-performance research client running natively at 60fps across iOS, Android, Windows Desktop, macOS, and Web.