Automatic extraction of angiogenesis bioprocess from text

Area of Science:

• Biomedical Informatics
• Computational Biology
• Drug Discovery

Background:

• Understanding biological processes (bioprocesses) is vital for drug design.
• Bioprocesses are challenging to identify in literature due to varied textual expressions and their composition of bioevents.
• Current methods struggle to capture bioprocesses referred to as bioevents.

Purpose of the Study:

• To develop and evaluate methods for identifying bioprocess terms and events in scientific literature.
• To address the challenge of capturing bioprocesses, particularly those described as bioevents.
• To improve information retrieval for drug design and discovery.

Main Methods:

• Development of a gold standard corpus annotated for angiogenesis (new blood vessel growth) terms and events.
• Utilizing domain-specific vocabularies, a manually annotated corpus, and unstructured domain-specific documents.
• Employing a supervised machine-learning model for bioprocess and bioevent identification.

Main Results:

• Over 36% of text expressions referring to angiogenesis were identified as events.
• The supervised machine-learning model demonstrated superior precision, recall, and F1 scores.
• Alternative methods provided reasonable performance with lower development costs.

Conclusions:

• Effective methods for identifying bioprocesses and bioevents in text have been presented.
• A supervised machine-learning approach offers the highest accuracy for this task.
• The developed resources (corpus, vocabularies) facilitate further research in bioprocess identification.