The gene normalization task in BioCreative III

Abstract

BackgroundWe report the Gene Normalization (GN) challenge in BioCreative III where participating teams were asked to return a ranked list of identifiers of the genes detected in full-text articles. For training, 32 fully and 500 partially annotated articles were prepared. A total of 507 articles were selected as the test set. Due to the high annotation cost, it was not feasible to obtain gold-standard human annotations for all test articles. Instead, we developed an Expectation Maximization (EM) algorithm approach for choosing a small number of test articles for manual annotation that were most capable of differentiating team performance. Moreover, the same algorithm was subsequently used for inferring ground truth based solely on team submissions. We report team performance on both gold standard and inferred ground truth using a newly proposed metric called Threshold Average Precision (TAP-k).ResultsWe received a total of 37 runs from 14 different teams for the task. When evaluated using the gold-standard annotations of the 50 articles, the highest TAP-k scores were 0.3297 (k=5), 0.3538 (k=10), and 0.3535 (k=20), respectively. Higher TAP-k scores of 0.4916 (k=5, 10, 20) were observed when evaluated using the inferred ground truth over the full test set. When combining team results using machine learning, the best composite system achieved TAP-k scores of 0.3707 (k=5), 0.4311 (k=10), and 0.4477 (k=20) on the gold standard, representing improvements of 12.4%, 21.8%, and 26.6% over the best team results, respectively.ConclusionsBy using full text and being species non-specific, the GN task in BioCreative III has moved closer to a real literature curation task than similar tasks in the past and presents additional challenges for the text mining community, as revealed in the overall team results. By evaluating teams using the gold standard, we show that the EM algorithm allows team submissions to be differentiated while keeping the manual annotation effort feasible. Using the inferred ground truth we show measures of comparative performance between teams. Finally, by comparing team rankings on gold standard vs. inferred ground truth, we further demonstrate that the inferred ground truth is as effective as the gold standard for detecting good team performance.

Normalising Medical Concepts in Social Media Texts by Learning Semantic Representation

“…(Tsuruoka et al, 2007;Ristad and Yianilos, 1998;Lu et al, 2011;McCallum et al, 2012). For example, McCallum et al (2012) used conditional random field to learn edit distance between phrases.…”

“…Traditional approaches, e.g. (Ristad and Yianilos, 1998;Aronson, 2001;Lu et al, 2011;McCallum et al, 2012), used proximity matching or heuristic string matching rules based on dictionary lookup when mapping texts to medical concepts. For example, Ristad and Yianilos (1998) incorporated edit-distance when mapping similar texts.…”

Knowledge-based bioinformatics for the study of mammalian oocytes

“…pathways, proteins, etc) and extracting the relationships with natural language processing methods. Other text mining software suites specifically designed for biomedical data analytics are reviewed in (Lu et al, 2011).…”

Gene Set to Diseases (GS2D): disease enrichment analysis on human gene sets with literature data

“…Automated text mining of these data has been already applied to derive gene-disease associations (e.g. [12,13]); however, these approaches require recognizing gene names using automated text mining methods that suffer from low accuracy [14,15]. Another way to extract gene-disease associations from the literature is to integrate the numerous manually curated annotations of PubMed citations for genes or diseases.…”