Scalable Knowledge Extraction from Biomedical Literature for Drug Discovery Applications

Biomedical literature has been growing exponentially, and thus, this has provided opportunities and challenges in the discovery of new drugs. As much as there is a lot of data contained in the published research, the extraction of meaningful knowledge by manual means is becoming impossible. The research in drug development requires the identification of the crucial biological entities and their relationships, which can be done with an automated, scalable strategy. The most difficult part is how to organize unstructured textual information to make it show drug-target interaction, illness causes, and treatment choices. Manual curation is unproductive and subject to error, and the continuous flow of new publications discourages researchers, making them unable to make evidence-based decisions in time. It presents BioNER-RE: a system that is a scalable approach that integrates Biomedical Named Entity Recognition (BioNER) and Relation Extraction (RE) on transformer-based models, including BioBERT. RE creates functional and mechanistic links between entities, whereas BioNER-RE identifies drugs, genes, proteins, and diseases. Querying a structured knowledge graph with extracted data can reveal new insights into medication repositioning and target finding. BioNER-RE detects entities and relationships in biomedical corpora, such as PubMed, more effectively than rule-based and keyword-matching approaches. Case studies support therapeutic discoveries by suggesting drug-target interactions. Ultimately, for data-driven and rapid drug discovery, BioNER-RE offers a reliable and adaptable biomedical literature mining system. This strategy can reduce research time, facilitate evidence-based judgments, and expedite the discovery of novel treatment options.