Distant Supervision for Cancer Pathway Extraction from Text

Pacific Symposium on Biocomputing, 2015 |

Biological pathways are central to understanding complex diseases such as cancer. The majority of this knowledge is scattered in the vast and rapidly growing research literature. To automate knowledge extraction, machine learning approaches typically require annotated examples, which are expensive and time-consuming to acquire. Recently, there has been increasing interest in leveraging databases for distant supervision in knowledge extraction, but existing applications focus almost exclusively
on newswire domains. In this paper, we present the first attempt to formulate the distant supervision problem for pathway extraction and apply a state-of-the-art method to extracting pathway interactions from PubMed abstracts. Experiments show that distant supervision can effectively compensate for the lack of annotation, attaining an accuracy approaching supervised results. From 22 million PubMed abstracts, we extracted 1.5 million pathway interactions at a precision of 25%. More than 10% of interactions are mentioned in the context of one or more cancer types, analysis of which yields interesting insights.