Full text
| |
| Author(s): |
Helder Ken Shimo
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Ribeirão Preto. |
| Institution: | Universidade de São Paulo (USP). Instituto de Matemática e Estatística (IME/SBI) |
| Defense date: | 2012-07-17 |
| Examining board members: |
Renato Tinós;
Paulo Sérgio Lopes de Oliveira;
Fernando José von Zuben
|
| Advisor: | Renato Tinós |
| Abstract | |
Many real world problems can be described as optimization problems. In bioinformatics in special, there is multiple sequence alignment, filogeny and RNA and Protein structure prediction, among others. Population based metaheuristics are techniques based in the interaction of a set of candidate solutions as elements of a population. Its use is specially interesting in optimization problems where there is little or no knowledge of the search space. The objective of this work is to study the use of self-organization of population in an artificial imune system for use in the docking problem, considered a complex multimodal optimization problem. The artificial imunme system is a type of population based methaheuristics inspired in the microevolution of the adaptive immune system of complex organisms. Candidate solutions represent cells of the immune system adapting its antibodies to eliminate a pathogen. The development of the algorithm was based in the opt-aiNet, based in the principles of clonal selection and affinity maturation for function optimization. Additionally, the opt-aiNet, inspired in theories of immune network, makes a suppression stage to eliminate similiar solutions and control diversity. This stage is computationally expensive as it calculates the distance between every possible pair of cells (solutions) eliminating those closer than a threshold. This work proposes a self-organized suppression algorithm inspired in the self-organized criticality, looking to minimize the influence of parameter selection and complexity of the suppression stage in opt-aiNet. The proposed algorithm was tested in a set of well-known functions in the evolutionary computation community. The results were compared to those of an implementation of the opt-aiNet. In addition, we proposed a mutation operator with q-Gaussian distribution for the artificial immune systems. The algorithm was then applied in the rigid protein docking problem based in surface complementarity and colision avoidance. The results were compared with a genetic algorithm and achieved a better performance. (AU) | |
| FAPESP's process: | 09/12944-8 - Self-Organization of the Population in Artificial Immune Systems |
| Grantee: | Helder Ken Shimo |
| Support Opportunities: | Scholarships in Brazil - Master |