Extraction of gene-disease association from literature using BioBERT

With the rapid growth of biomedical literatures, there are a large amount of bio-text data to be exploited. A wealth of knowledge concerning diseases associated with genes is present in those bio-text which is important for studies like drug-target discovery, even provide personalized medical treatment for different patients' genome conditions. BioBERT as a pre-trained BERT model with large-scale biomedical corpora, was proved has a great performance over other pre-trained language models on biomedical datasets. To make the use of a large amount of bio-text, in this paper we provide a good practice that use BioBERT to extract the gene-disease associations from bio-text, and it achieved an overall F-score of 79.98%. Hoping to inspire researchers in the biomedical field of natural language processing and be able to make applications in related fields to solve the problems encountered in the research.