Automated generation of modular of gene descriptions
Year: 2019
Type: other
Abstract: Summarizing gene function into short human-readable text from an ever-growing literature corpus is a time- and labor-intensive task. In order to automate the writing of a gene description, we developed an algorithm that uses curated gene data provided by member databases of the Alliance of Genome Research (Alliance; www.alliancegenome.org). Data includes gene associations to Gene Ontology (GO) terms [1], Disease Ontology (DO) terms [2] and human orthologs. This project has resulted in modular and standardized gene descriptions for the seven species at the Alliance portal (yeast, worm, fly, zebrafish, mouse, rat and human), including those that lacked them, for purposes of display and use at the Alliance and at the relevant Model Organism Databases, thus facilitating discoverability and interspecies comparison.The problem of generating a text summary from multiple ontology terms annotated to a gene was formulated as a set-covering problem and solved with a ‘greedy’ algorithm, which is the best known approach to solve the problem in polynomial time. The algorithm is designed to balance readability of a gene description with the amount of information it provides. The software is provided as an open source Python package and integrated into the build process of the Alliance website; it can be downloaded and integrated into other pipelines wherever text descriptions are needed. For example, the software is used in a custom WormBase pipeline to generate gene descriptions for C. elegans and nine other nematode species. We are currently refining the algorithm by methods that include weighting ontology terms by information content and using data-type specific rules.[1] The Gene Ontology Consortium. Expansion of the Gene Ontology knowledgebase and resources. Nucleic Acids Res. 2017 Jan 4;45(D1):D331-D338.[2] Schriml LM et al., Human Disease Ontology 2018 update: classification, content and workflow expansion. Nucleic Acids Res. 2018 Nov 8.
Authors: Ranjana Kishore, Valerio Arnaboldi, Ceri E. Van Slyke, Stacia R. Engel, Juancarlos Chan, Susan M. Bello, Robert S. Nash, Jose M. Urbano, Mary E. Dolan, Mary Shimoyama, Paul W. Sternberg, The Alliance of Genome Resources
Institution: California Institute of Technology
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W2951509849