Triage of documents containing protein interactions affected by mutations using an NLP based machine learning...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomics

Background:

• Understanding how mutations affect protein-protein interactions (PPIs) is crucial for disease mechanism elucidation and therapeutic development.
• Existing literature contains valuable information on mutation-affected PPIs, but manual extraction is time-consuming and challenging.
• Automated methods are needed to efficiently mine this complex biological data.

Purpose of the Study:

• To develop and evaluate a natural language processing (NLP) based machine learning system for extracting information on protein-protein interactions affected by mutations from scientific texts.
• To identify relevant documents (abstracts or full-text paragraphs) containing mutation-impacted PPIs.

Main Methods:

• Utilized a machine learning approach powered by advanced NLP techniques.
• Engineered a comprehensive set of features, including those derived from pre-trained word embeddings.
• Applied the system to the Document Triage Task within the BioCreative VI Precision Medicine Track.

Main Results:

• The developed system demonstrated satisfactory performance in identifying relevant scientific literature.
• Achieved the highest recall and a competitive F1-score in the BioCreative VI Precision Medicine Track's Document Triage Task.
• The NLP-based approach effectively extracts mutation-specific PPI information.

Conclusions:

• The system's performance suggests its utility in complementing manual annotation efforts.
• The machine learning framework and feature engineering can benefit future research in biological text mining.
• This approach facilitates the advancement of precision medicine through automated literature analysis.