OPA2Vec: combining formal and informal content of biomedical ontologies to improve similarity-based prediction

Abstract

Motivation: Ontologies are widely used in biology for data annotation, integration, and analysis. In addition to formally structured axioms, ontologies contain meta-data in the form of annotation axioms which provide valuable pieces of information that characterize ontology classes. Annotation axioms commonly used in ontologies include class labels, descriptions, or synonyms. Despite being a rich source of semantic information, the ontology meta-data are generally unexploited by ontology-based analysis methods such as semantic similarity measures. Results: We propose a novel method, OPA2Vec, to generate vector representations of biological entities in ontologies by combining formal ontology axioms and annotation axioms from the ontology metadata. We apply a Word2Vec model that has been pre-trained on either a corpus or abstracts or full-text articles to produce feature vectors from our collected data. We validate our method in two different ways: first, we use the obtained vector representations of proteins in a similarity measure to predict proteinprotein interaction on two different datasets. Second, we evaluate our method on predicting gene-disease associations based on phenotype similarity by generating vector representations of genes and diseases using a phenotype ontology, and applying the obtained vectors to predict gene-disease associations using mouse model phenotypes. We demonstrate that OPA2Vec significantly outperforms existing methods for predicting gene-disease associations. Using evidence from mouse models, we apply OPA2Vec to identify candidate genes for several thousand rare and orphan diseases. OPA2Vec can be used to produce vector representations of any biomedical entity given any type of biomedical ontology.

Phenotypic, functional and taxonomic features predict host-pathogen interactions

“…These ontologies contain formalized biological background knowledge (Hoehndorf et al, 2015b). Using the information in ontologies as background knowledge during feature generation has the potential to significantly improve the performance of these features in machine learning and predictive analyses (Smaili et al, 2018).…”

“…We use the OPA2Vec (Smaili et al, 2018) method to generate features from annotations of pathogen taxa and human proteins while using the ontologies that are used to express them as background knowledge.…”

“…For feature generation, we used OPA2Vec (Smaili et al, 2018), a tool that generates feature vectors from entities and their annotation from ontologies. OPA2vec first generates a corpus consisting of relevant annotations of pathogens and proteins (i.e., HP, MP, GO, and NCBI Taxonomy classes), the complete set of class axioms and the metadata of the ontologies.…”

“…We systematically compare phenotypes arising from infection with a pathogen to phenotypes associated with human genes; we further relate phenotypes to protein functions and compare pathogen-associated phenotypes with human protein functions. Relying on recent progress in deep learning with structured and semantic data (Smaili et al, 2018), we construct a model that combines this information as well as structured information about taxonomic relations between pathogens to predict host-pathogen interactions. We demonstrate that our model can not only accurately predict the interacting protein partners for pathogens with phenotypic information, but also their relatives without phenotypes by aggregating knowledge through the taxonomy.…”

Ontology-based prediction of cancer driver genes

“…The goal of identifying cancer 17 drivers may be achieved at the level of gene, protein or pathways, and multiple 18 approaches have been attempted to date [4]. There is no gold standard against which 19 the success of an algorithm can be measured, although the Cancer Gene Census 20 approaches a "gold standard" most closely with an expert-curated dataset of 21 cancer-associated genes and mutations [5]. 22 Investigators have increasingly relied on taking a consensus of multiple methods and, 23 where possible, attempted to experimentally verify driver gene status in cellular or 24 whole organism systems [6].…”