AI summaryⓘ
The authors studied how to best connect biomedical scientific text with structured knowledge graphs (KGs), which are like detailed databases of facts. They froze existing text and KG models and trained a simple mapping between them to compare different design choices. Using a new dataset linking chemical-gene interactions to supporting articles, they tested how well texts and KG triples could be matched in two directions. They found that how the KG triple is represented and the direction of training matter most, while the choice of text encoder and sampling methods are less important. Overall, the simplest approach—projecting text into the KG space using a linear projection on combined triple parts—worked best.
Biomedical knowledge graphContrastive learningText encoderKG embeddingChemical-gene interactionsKnowledge alignmentProjection headHard-negative samplingComparative Toxicogenomics DatabaseInformation retrieval
Authors
Artem Bisliouk, Elizaveta Nosova, Heiko Paulheim, Andreea Iana, Rita T. Sousa
Abstract
Biomedical knowledge exists in two complementary but distinct forms: unstructured scientific literature and structured knowledge graphs (KGs). Aligning them is essential for knowledge grounding, evidence retrieval, and KG completion, yet existing methods do not explicitly align free-text evidence with KG triples. We present a unified framework for systematically studying design choices for aligning biomedical text and KGs. With a text encoder and a KG embedding model both frozen, we learn only a lightweight projection between their spaces via a contrastive objective. This enables a fair comparison across six design dimensions: text encoder, KG embedding model, projection head, triple composition, training direction, and hard-negatives sampling. We construct CTD-Align, a corpus of over 22K one-to-one tripledocument pairs linking chemical-gene interactions from the Comparative Toxicogenomics Database to supporting PubMed passages. We evaluate alignment on it in two retrieval settings: document-to-triple and triple-to-document. We find that the triple composition and the training direction (i.e., shared retrieval space) have the greatest impact, whereas the text encoder and hard-negatives sampling matter little. Overall, simple choices win: projecting text into the KG space with a linear head over concatenated subject, predicate, and object embeddings performs best. These findings establish lightweight contrastive alignment as an effective, practical foundation for bridging biomedical text and KGs.