Biomedical Information Extraction Pipeline to Identify Disease-Gene Interactions from PubMed Breast Cancer Literature

Many valuable insights and enormous amount of knowledge are buried in the unstructured data of biomedical literature. With the help of technological progression, organizations and drug manufacturers have invested in extracting major findings and insights from the vast amount of data. Extracting the drug names, diseases, genes is now easy due to technology innovation in Text Mining and Natural Language Processing. Entity recognition and relations extraction have helped numerous pharmacists reduce the amount of time in novel drugs production. In this article, the innovative text-mining tools were used to identify named entities and automatically extract thousands of disease-gene relations from 3000 publications from PubMed medical literature specified in Breast Cancer. Firstly, the bio-entities in the dataset were recognized by ScispaCy. After that, the relations between the extracted disease-gene entities were identified by the NeMo toolkit. The trained model provided a decent accuracy (F1 score: 80.03) in the drug-disease associations, which plays a significant role in developing novel drugs for incurable diseases.