Conference Papers
Permanent URI for this collectionhttps://idr.nitk.ac.in/handle/123456789/28506
Browse
3 results
Search Results
Item A phenomic approach to genetic algorithms for reconstruction of gene networks(2010) D'Souza, R.G.L.; Chandra Sekaran, K.C.; Kandasamy, A.Genetic algorithms require a fitness function to evaluate individuals in a population. The fitness function essentially captures the dependence of the phenotype on the genotype. In the Phenomic approach we represent the phenotype of each individual in a simulated environment where phenotypic interactions are enforced. In reconstruction type of problems, the model is reconstructed from the data that maps the input to the output. In the phenomic algorithm, we use this data to replace the fitness function. Thus we achieve survival-of-the- fittest without the need for a fitness function. Though limited to reconstruction type problems where such mapping data is available, this novel approach nonetheless overcomes the daunting task of providing the elusive fitness function, which has been a stumbling block so far to the widespread use of genetic algorithms. We present an algorithm called Integrated Pheneto-Genetic Algorithm (IPGA), wherein the genetic algorithm is used to process genotypic information and the phenomic algorithm is used to process phenotypic information, thereby providing a holistic approach which completes the evolutionary cycle. We apply this novel evolutionary algorithm to the problem of elucidation of gene networks from microarray data. The algorithm performs well and provides stable and accurate results when compared to some other existing algorithms. © 2010 Springer-Verlag Berlin Heidelberg.Item Inference of Gene Networks from Microarray Data through a Phenomic Approach(2010) D'Souza, R.G.L.; Chandra Sekaran, K.C.; Kandasamy, A.The reconstruction of gene networks is crucial to the understanding of cellular processes which are studied in Systems Biology. The success of computational methods of drug discovery and disease diagnosis is dependent upon our understanding of the biological basis of the interaction networks between the genes. Better modelling of biological processes and powerful evolutionary methods are proving to be a key factor in the solution of such problems. However, most of these methods are based on processing of genotypic information. We present an evolutionary algorithm for inferring gene networks from expression data using phenotypic interactions. The benefit of this is that we avoid the need for an explicit objective function in the optimization process. In order to realize this, we have implemented a method called as the Phenomic algorithm and validated it for stability and accuracy in the reconstruction of gene networks. © Springer-Verlag Berlin Heidelberg 2010.Item A phenomic algorithm for inference of gene networks using S-systems and memetic search(2012) D'Souza G, R.G.L.; Chandra Sekaran, K.C.; Kandasamy, A.In recent years, evolutionary methods have seen unprecedented success in elucidation of gene networks, especially from microarray data. We have implemented the Phenomic Algorithm which is an evolutionary method for inference of gene networks based on population dynamics. We have used S-systems to model gene interactions and applied memetic search to fine tune the parameters of the inferred networks. We have tested the novel algorithm on artificial gene expression datasets obtained from simulated gene networks. We have also compared the results to those obtained from two other similar algorithms. Results showed that the new method, which we call as Phenomic Algorithm with Memetic Search (PAMS), is an effective method for inference of gene networks. © 2012 ICST Institute for Computer Science, Social Informatics and Telecommunications Engineering.