Full text
| |
| Author(s): |
André Atanasio Maranhão Almeida
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Computação |
| Defense date: | 2013-02-21 |
| Examining board members: |
Zanoni Dias;
Guilherme Pimentel Telles;
Orlando Lee;
Felipe Rodrigues da Silva;
Maria Emilia Machado Telles Walter
|
| Advisor: | Zanoni Dias |
| Abstract | |
Sequence alignment is one the most important tasks of bioinformatics. It is a basic operation used for several procedures in that domain, such as sequence database searches, evolution effect visualization in an entire protein family, phylogenetic trees construction and preserved motifs identification. Sequences can be aligned in pairs and generate a pairwise alignment. Three or more sequences can also be simultaneously aligned and generate a multiple sequence alignment (MSA). MSAs could be used for pattern recognition for protein family characterization and secondary and tertiary protein structure prediction. Let l be the sequence length. The pairwise alignment takes time O(l2) to build an exact alignment. However, multiple sequence alignment is a NP-Hard problem. In this work, heuristic methods were developed for multiple protein sequence alignment. The main approaches and methods applied to the problem were studied and a series of aligners developed and evaluated. In a first moment 342 multiple aligners using the progressive approach were built. That is a largely used approach for MSA construction and is composed by three steps. In the first one a distance matrix is computed. Then, a guide tree is built based on the matrix and finally the MSA is constructed through pairwise alignments. The order to the pairwise alignments is defined by the tree. The developed aligners combine distinct methods applied to each of steps. Then, evaluations in the consistency based alignment context were performed. In that approach, a MSA is optimal when agree with the majority along all possible optimal pairwise alignments. MUMMALS is a known consistency based aligner. It was changed in this evaluation. The k-mer counting method was modified in two distinct ways. The k value and the compressed alphabet were ranged. In another evaluation, the k-mer counting method and guide tree construction method were replaced. In the last stage of the work, iterative algorithms were developed and evaluated. Those methods are characterized by other aligner's dependence. The other aligners generate an initial population and the iterative aligner performs a refinement procedure, which iteratively changes the alignments until the alignments quality are stabilized. Several evaluations were performed. However, a genetic algorithm for MSA construction stood out along this stage. In that aligner were added other approaches for multiple sequence alignment construction, such as block based, consensus based and template based. The first one was applied to initial population generation, the second one was used for a crossover operator creation and the third one defined a fitness function (AU) | |
| FAPESP's process: | 07/08020-0 - New approaches for the problem of multiple sequence alignment |
| Grantee: | André Atanasio Maranhão Almeida |
| Support Opportunities: | Scholarships in Brazil - Doctorate |