eGARD: Extracting associations between genomic anomalies and drug responses from text

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Personalized cancer treatment relies on identifying genomic anomalies and their therapeutic implications.
• Biomedical literature contains vast, yet inaccessible, information on biomarker-drug response relationships.
• Manual literature review is time-consuming and challenging for clinicians and researchers.

Purpose of the Study:

• To develop an automated system for extracting genomic anomaly-drug response associations from biomedical literature.
• To assist biocurators, researchers, and oncologists in keeping pace with rapidly growing scientific information.
• To facilitate the identification of biomarkers for personalized cancer therapy selection.

Main Methods:

• A natural language processing (NLP)-based text mining (TM) system named eGARD (extracting Genomic Anomalies association with Response to Drugs) was developed.
• The system analyzes MEDLINE abstracts, utilizing sentence syntax and textual features to identify relationships between genomic anomalies and drug response.
• Evaluation datasets from in-house annotations and PharmGKB were used to assess system performance.

Main Results:

• The eGARD system achieved high performance metrics: precision up to 0.95, recall up to 0.86, and F-measure up to 0.90.
• The system successfully extracts information relevant to the clinical utility of genomic anomalies.
• Extracted data includes confidence levels to aid in prioritizing curation efforts.

Conclusions:

• The eGARD system offers an efficient solution for extracting critical information from biomedical literature.
• This NLP-based tool can significantly speed up the review process, enabling faster clinical research.
• eGARD empowers researchers to stratify patients for targeted therapies and generate hypotheses for new clinical trials.