A Centrality Based Multi-Objective Disease-Gene Association Approach Using Genetic Algorithms
The Disease Gene Association Problem (DGAP) is a bioinformatics problem in which genes are ranked with respect to how involved they are in the presentation of a particular disease. Previous approaches have shown the strength of both Monte Carlo and evolutionary computation (EC) based techniques. Typically these past approaches improve ranking measures, develop new gene relation definitions, or implement more complex EC systems. This thesis presents a hybrid approach which implements a multi-objective genetic algorithm, where input consists of centrality measures based on various relational biological evidence types merged into a complex network. In an effort to explore the effectiveness of the technique compared to past work, multiple objective settings and different EC parameters are studied including the development of a new exchange methodology, safe dealer-based (SDB) crossover. Successful results with respect to breast cancer and Parkinson's disease compared to previous EC techniques and popular known databases are shown. In addition, the newly developed methodology is also successfully applied to Alzheimer’s, further demonstrating the flexibility of the technique. Across all three cases studies the strongest results were produced by the shortest path-based measures stress and betweenness in a single objective parameter setting. When used in conjunction in a multi-objective environment, competitive results were also obtained but fell short of the single objective settings studied as part of this work. Lastly, while SDB crossover fell short of expectations on breast cancer and Parkinson's, it achieved the best results when applied to Alzheimer’s, illustrating the potential of the technique for future study.