Predicting candidate genes from phenotypes, functions, and anatomical site of expression

Abstract

Motivation: Over the past years, many computational methods have been developed to incorporate information about phenotypes for disease gene prioritization task. These methods generally compute the similarity between a patient's phenotypes and a database of gene-phenotype to find the most phenotypically similar match. The main limitation in these methods is their reliance on knowledge about phenotypes associated with particular genes, which is not complete in humans as well as in many model organisms such as the mouse and fish. Information about functions of gene products and anatomical site of gene expression is available for more genes and can also be related to phenotypes through ontologies and machine learning models. Results: We developed a novel graph-based machine learning method for biomedical ontologies which is able to exploit axioms in ontologies and other graph-structured data. Using our machine learning method, we embed genes based on their associated phenotypes, functions of the gene products, and anatomical location of gene expression. We then develop a machine learning model to predict gene-disease associations based on the associations between genes and multiple biomedical ontologies, and this model significantly improves over state of the art methods. Furthermore, we extend phenotype-based gene prioritization methods significantly to all genes which are associated with phenotypes, functions, or site of expression. Availability: Software and data are available at https://github. com/bio-ontology-research-group/DL2Vec.

DTI-Voodoo: machine learning over interaction networks and ontology-based background knowledge predicts drug–target interactions

“…As phenotypes and functions are encoded through ontologies, we use DL2Vec (Chen et al, 2020) to obtain ontology based representations for use as top-down features. DL2vec constructs a graph by introducing nodes for each ontology class and edges for ontology axioms, followed by random walks starting from each node in the graph.…”

“…Additionally, the generated features may be used for other tasks. We follow the results of DL2vec (Chen et al, 2020) and use σ := LeakyReLU as activation function which leads to improved performance compared to other activation functions.…”

“…We developed DTI-Voodoo as a method for predicting DTIs. We use an ontology-based machine learning method (Chen et al, 2020) to encode phenotypic consequences of DTIs and deep learning methods to encode molecular features. We combine both using a protein interaction network which we exploit with the aid of a graph neural network.…”

DeepViral: infectious disease phenotypes improve prediction of novel virus–host interactions

“…To generate feature embeddings, we used DL2Vec (Chen et al, 2020), a recent method for learning features for entities (in our case, the human proteins and viruses) from their associations to ontological classes. DL2Vec first converted the ontologies and entity associations into a graph, with the classes and entities as the nodes and the associations and ontology axioms as the edges.…”

“…To generate feature embeddings for human proteins and virus taxa, we applied a recent representation learning method DL2Vec (Chen et al, 2020), which learned feature embeddings for entities based on their annotations to ontology classes (see Section 2.2). DL2Vec takes two types of inputs: the associations of the entities with ontology classes (e.g., human proteins and their functions), and the ontologies themselves.…”