Case Study: Wisecube enhances biomedical knowledge graphs with Lettria’s Text-to-Graph pipeline

Enhancing Biomedical Knowledge Graphs with Lettria’s Text-to-Graph

Wisecube, a company specializing in AI solutions for the life sciences, faced the challenge of expanding its biomedical knowledge graph. They needed to integrate complex, unstructured data from PubMed abstracts, which was difficult due to the dense scientific text and the massive scale of the required Biolink ontology. Their goal was to accurately extract entities and relationships and align them with their existing Wikidata-based graph. Lettria provided its Text-to-Graph pipeline to address this.

Lettria's solution involved modularizing the large Biolink ontology into manageable segments for efficient processing by a large language model. This enabled the accurate extraction of thousands of triples from biomedical text. The extracted data was aligned with Wikidata entities and seamlessly integrated into Wisecube's knowledge graph stored on AWS Neptune. The results were significant, with Lettria processing 108 GB of text from 350,000 articles to extract 3,500 triples, enhancing the graph with high-quality biomedical data and ensuring scalable, accurate integration.

Wisecube