Research On Informative SNP Selection Method Based On Intelligent Algorithm

Single Nucleotide Polymorphism (SNP) is a DNA set of polymorphism based onthe single nucleotide variations at the genomic level. Although analyzing all SNP canreach high association power, its happlotyping cost is huge. Selecting informativeSNPs can reduce cost while maintaining association. Currently, there are manymethods for informative SNP selection. However, they also have some drawbacksincluding: high time complexity, the big number of informative SNP, low predictionaccuracy or loss much information for genetic association. Thus, we propose a novelmethod based on intelligent algorithm in order to improve performance and meet theneed of association study. The creativities and contribution are discussed in detail asfollows:In first, the methods to select tagSNPs are introduced, including several mainideas. Moreover, these methods are compared to show their advantages, disadvantagesand their scope of application. And, the basic idea of methods based on predictionaccuracy is detailed. Then, a new method based on ant colony algorithm is introducedto construct subsets of informative SNPs. In order to reduce time complexity, wepropose a method for constructing subsets based on ant colony algorithm, and thendesign effective heuristic function based on prediction, ant decision and adaptivepheromone updating mechanism. More importantly, by making full use of the multipleoutput of artificial neural network and designing valid output function based distancemodel, we propose the ANN to predict non-tag SNPs, elevating the predictionaccuracy and speed. Finally, in order to facilitate that other biological informationprocessing researchers effectively use this method, we design and implement thissoftware based on VC6.0platform. This new method is implemented in this software.Then, we use it to process various simulated datasets and real datasets, and comparewith other method. Experiments prove that the algorithm largely solves the problemof local optimization and the accuracy of the prediction results is further improved.