Limits...
Evolutionary algorithms for the selection of single nucleotide polymorphisms.

Hubley RM, Zitzler E, Roach JC - BMC Bioinformatics (2003)

Bottom Line: The choice of subset is influenced by many factors, including estimated or known reliability of the SNP, biochemical factors, intellectual property, cost, and effectiveness of the subset for mapping genes or identifying disease loci.They provide flexibility with respect to the problem formulation if a problem description evolves or changes.Results are produced as a trade-off front, allowing the user to make informed decisions when prioritizing factors.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute for Systems Biology, Seattle, WA, USA. rhubley@systemsbiology.org

ABSTRACT

Background: Large databases of single nucleotide polymorphisms (SNPs) are available for use in genomics studies. Typically, investigators must choose a subset of SNPs from these databases to employ in their studies. The choice of subset is influenced by many factors, including estimated or known reliability of the SNP, biochemical factors, intellectual property, cost, and effectiveness of the subset for mapping genes or identifying disease loci. We present an evolutionary algorithm for multiobjective SNP selection.

Results: We implemented a modified version of the Strength-Pareto Evolutionary Algorithm (SPEA2) in Java. Our implementation, Multiobjective Analyzer for Genetic Marker Acquisition (MAGMA), approximates the set of optimal trade-off solutions for large problems in minutes. This set is very useful for the design of large studies, including those oriented towards disease identification, genetic mapping, population studies, and haplotype-block elucidation.

Conclusion: Evolutionary algorithms are particularly suited for optimization problems that involve multiple objectives and a complex search space on which exact methods such as exhaustive enumeration cannot be applied. They provide flexibility with respect to the problem formulation if a problem description evolves or changes. Results are produced as a trade-off front, allowing the user to make informed decisions when prioritizing factors. MAGMA is open source and available at http://snp-magma.sourceforge.net. Evolutionary algorithms are well suited for many other applications in genomics.

Show MeSH
SPEA2. A cartoon of the SPEA2 algorithm, operating on a population size of eight with an archive of four. The algorithm continues for a fixed number of generations, and then outputs non-dominated solutions. Details of each of the steps are described in the text.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC183839&req=5

Figure 2: SPEA2. A cartoon of the SPEA2 algorithm, operating on a population size of eight with an archive of four. The algorithm continues for a fixed number of generations, and then outputs non-dominated solutions. Details of each of the steps are described in the text.

Mentions: An evolutionary algorithm is initialized with a seed population of many random, or possibly non-random, solutions. The algorithm then steps through a number of iterations (Figure 2). During each iteration, the best solutions from the previous iteration are allowed to mutate and recombine. These new solutions then compete with each other, and with the previous best solutions. Good solutions are retained; poor solutions are rejected. Evolutionary algorithms work well when the operators for mutation and recombination are likely to efficiently explore the solution space. If good solutions to a problem are not reachable by mutation or recombination of other solutions, then an evolutionary algorithm will not be able to find these solutions. However, it is unlikely that any algorithm short of exhaustive search would be able to find such solutions.


Evolutionary algorithms for the selection of single nucleotide polymorphisms.

Hubley RM, Zitzler E, Roach JC - BMC Bioinformatics (2003)

SPEA2. A cartoon of the SPEA2 algorithm, operating on a population size of eight with an archive of four. The algorithm continues for a fixed number of generations, and then outputs non-dominated solutions. Details of each of the steps are described in the text.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC183839&req=5

Figure 2: SPEA2. A cartoon of the SPEA2 algorithm, operating on a population size of eight with an archive of four. The algorithm continues for a fixed number of generations, and then outputs non-dominated solutions. Details of each of the steps are described in the text.
Mentions: An evolutionary algorithm is initialized with a seed population of many random, or possibly non-random, solutions. The algorithm then steps through a number of iterations (Figure 2). During each iteration, the best solutions from the previous iteration are allowed to mutate and recombine. These new solutions then compete with each other, and with the previous best solutions. Good solutions are retained; poor solutions are rejected. Evolutionary algorithms work well when the operators for mutation and recombination are likely to efficiently explore the solution space. If good solutions to a problem are not reachable by mutation or recombination of other solutions, then an evolutionary algorithm will not be able to find these solutions. However, it is unlikely that any algorithm short of exhaustive search would be able to find such solutions.

Bottom Line: The choice of subset is influenced by many factors, including estimated or known reliability of the SNP, biochemical factors, intellectual property, cost, and effectiveness of the subset for mapping genes or identifying disease loci.They provide flexibility with respect to the problem formulation if a problem description evolves or changes.Results are produced as a trade-off front, allowing the user to make informed decisions when prioritizing factors.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute for Systems Biology, Seattle, WA, USA. rhubley@systemsbiology.org

ABSTRACT

Background: Large databases of single nucleotide polymorphisms (SNPs) are available for use in genomics studies. Typically, investigators must choose a subset of SNPs from these databases to employ in their studies. The choice of subset is influenced by many factors, including estimated or known reliability of the SNP, biochemical factors, intellectual property, cost, and effectiveness of the subset for mapping genes or identifying disease loci. We present an evolutionary algorithm for multiobjective SNP selection.

Results: We implemented a modified version of the Strength-Pareto Evolutionary Algorithm (SPEA2) in Java. Our implementation, Multiobjective Analyzer for Genetic Marker Acquisition (MAGMA), approximates the set of optimal trade-off solutions for large problems in minutes. This set is very useful for the design of large studies, including those oriented towards disease identification, genetic mapping, population studies, and haplotype-block elucidation.

Conclusion: Evolutionary algorithms are particularly suited for optimization problems that involve multiple objectives and a complex search space on which exact methods such as exhaustive enumeration cannot be applied. They provide flexibility with respect to the problem formulation if a problem description evolves or changes. Results are produced as a trade-off front, allowing the user to make informed decisions when prioritizing factors. MAGMA is open source and available at http://snp-magma.sourceforge.net. Evolutionary algorithms are well suited for many other applications in genomics.

Show MeSH