Overview of the Cancer Genetics and Pathway Curation tasks of BioNLP Shared Task 2013

Area of Science:

• Biomedical Natural Language Processing
• Bioinformatics
• Computational Biology

Background:

• BioNLP Shared Task events since 2009 have driven automatic information extraction from biomedical literature.
• Introduced in 2013, the Cancer Genetics (CG) and Pathway Curation (PC) tasks address specific challenges in biomedical text mining.

Purpose of the Study:

• To present the Cancer Genetics (CG) and Pathway Curation (PC) tasks from the BioNLP Shared Task 2013.
• To evaluate the performance of event extraction methods on these novel tasks.

Main Methods:

• Six groups participated in the CG task, and two in the PC task.
• A variety of established and new information extraction approaches were applied.

Main Results:

• The top systems achieved 55% F-score on the CG task and 53% F-score on the PC task.
• Performance levels are comparable to similar previous biomedical event extraction tasks.

Conclusions:

• Existing event extraction technology can be generalized to the CG and PC task settings.
• These methods can aid in constructing knowledge bases for cancer molecular mechanisms and curating biomolecular pathway models.
• The CG and PC tasks remain open challenges with available data and resources.