Exploring Diseases Relationships: An Ontology-Based Spreading Activation Approach

Introduction

In the era of extraordinary large-scale data analysis empowered by the vast amount of biological information produced every day, biomedical research plays a key role. Therefore, there is a great demand to explore effective methods for biomedical data integration and knowledge management (Goble & Stevens, 2008). The semantic web gives the meaning of knowledge making software systems perform intelligent tasks instead of users (Berners-Lee, Hendler, & Lassila, 2001) using a semantic knowledgebase called Ontology. The ontology is a repository in which information is organized and used in semantic-based applications in order to support reasoning process such as detecting inconsistencies, identifying subsumption relationships, and instance classification (Said Fathalla and Christoph Lange, 2018a; Shadbolt, Hall, & Berners-Lee, 2006). Developing an efficient algorithm for traversing large ontologies is a key challenge for many semantic-based systems. This chapter introduces an ontology-based spreading activation approach, called SAOO (Spreading Activation over Ontology), in order to explore the relationship between two human diseases, i.e., the relatedness. If a relationship exists, the set of diseases connecting them is provided. For instance, is there a relation between “Pedophilia” and “Voyeurism”? If so, what is the relatedness? An interesting feature of finding relations between diseases is the treatment of the real cause of a disease, which might be because of the existence of another disease (Fathalla & Kannot, 2017a). For instance, Gallstones disease, bile forms solid particles (stones) in the gallbladder, might be caused by “Hemolytic Anemia” disease so crushing gallstones is not a solution or treatment because the stones will develop again (Trotman, Bernstein, Bove, & Wirth, 1980). Therefore, the objectives of detecting the relatedness between diseases are:

• •

  Causality: Possibly, the existence of disease might result in another disease to occur. For example, Hereditary Spherocytosis diseases are an autosomal preponderant anomaly of erythrocytes that causes gallstones. Pigmented gallstones occur in approximately half of the untreated patients,

• •

  Complications of Diseases: A disease can increase the complications of another disease, e.g., Diabetes and HCV hepatitis,

• •

  Treatments Prescription: A treatment prescription for a disease can change when there is a relationship between this disease and another one, e.g., HCV hepatitis and hypertension.

The methodology of the proposed approach is divided into two phases: semantic matching and diseases relatedness detection. In the semantic matching phase, user-submitted diseases are semantically identified in the ontology graph using the proposed matching algorithm. In the diseases relatedness detection phase, the URIs of these diseases are passed to the relatedness detector which in turn detects the relationship using the bidirectional spreading activation algorithm on the disease ontology and returns the set of diseases connecting them. The idea behind the use of bidirectional search is to find a path starting from the source node, i.e., the first disease, and the goal node, i.e., the second disease until an intersection occurs. In other words, it runs two simultaneous searches: forward from one node and backward from the other. When the two search frontiers meet, the algorithm will reconstruct a path, represent the set of diseases connect them. The knowledge base here is the disease ontology and Semantic Web Rule Language (SWRL) rules. In order to support the inference process, a set of SWRL rules is defined. From these rules, new relationships could be discovered from instance data, which did not explicitly exist in the ontology. Usage of SWRL rules extends the expressivity of OWL semantics (Tsarkov, Riazanov, Bechhofer, & Horrocks, 2004).