Gene Prioritization Using Semantic Similarity

Abstract

Molecular biologists and medical geneticists have to analyze large lists of genes and genomic variants to select the best candidates linked to the phenotype of interest. This task is very challenging therefore computational algorithms are used to prioritize candidate genes. Genes are prioritized based on evidence in favour of their relationship to the phenotype. A technique among other used in this process is measuring semantic similarity (SS) between functional annotations of genes. For SS to be applicable the functional annotations (knowledge about the domain) have to be in a form of hierarchically organized ontology. In this article we outline a computational process of estimating semantic similarity of phenotypic descriptions and gene annotations in Human Phenotype Ontology (HPO) and Gene Ontology. An example of a possible process of differential diagnosis applying semantic similarity to compare HPO phenotypes is presented. A semantic content of Gene Ontology annotations is highlighted. An example of comparison of meanings of individual GO annotations is presented and extended to a comparison of functional similarity of gene products.