miCDER: a context-aware transformer model for joint miRNA-disease entity and multi-level regulatory relation...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomics

Background:

• MicroRNA (miRNA) dysregulation is implicated in various human diseases.
• Existing methods struggle with standardized, fine-grained miRNA-disease interaction data and multi-level regulatory analysis.

Purpose of the Study:

• To develop a robust model for extracting detailed miRNA-disease regulatory interactions.
• To construct a comprehensive knowledge graph of miRNA-disease relationships.

Main Methods:

• Construction of a fine-grained, multi-level annotated biomedical dataset.
• Development of miCDER, a transformer-based model for joint entity and relation extraction.
• Utilizing contextual encoding and inter-span transformer attention.

Main Results:

• miCDER achieved state-of-the-art performance: NER F1=87.34%, RE F1=77.28%, outperforming baselines like SpERT.
• Demonstrated generalizability on CoNLL04 and ADE datasets with competitive and superior results, respectively.
• Extracted 93,221 regulatory triplets, constructing the MAAD-HCD-KG, including 1,735 novel targeting relations.

Conclusions:

• The miCDER model effectively extracts fine-grained regulatory information from biomedical texts.
• The generated MAAD-HCD-KG knowledge graph advances research on miRNA-related diseases.