Ranking of disease gene associations from large corpora of scientific publications
Year: 2015
Type: dissertation
Abstract: The extraction of disease-gene associations from biomedical publications is a widely inves-\ntigated field of research. In previous work, a frequent method was to implement natural\nlanguage processing tools that use semantic information to find such associations. How-\never, most of these approaches are restricted to single documents. Retrieval systems that\npredict novel associations across various documents often lack the ability to deal with the\nhuge amount of resulting candidates. In this work, we present a system that aggregates\ninformation from a large corpora of scientific abstracts. This information is used to build a\ncomprehensive gene-interaction network, which is then used to predict novel disease-gene\nassociations. We tackle the problem of candidate reduction by integrating two separate\nmachine learning methods. We train a support vector machine to classify genes as disease\nrelated or not and a support vector regression model to rank gene-candidates according to\ntheir importance to a specific disease. Thereto, we make use of approved methods and ex-\ntend them by a novel investigation of the gene-interaction network. In a model-evaluation\non two gold standards as well as in a case-study in cooperation with biomedical experts,\nit is shown that the proposed methods are able to extract disease-gene-associations from\nsingle documents and discover disease-related candidates across multiple documents.
Author: Hendrik ter Horst
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W2221029491
