Dhaka University Journal of Applied Science & Engineering

With the advent of high throughput DNA microarray technology, it is now possible to measure the nRNA expression levels of thousands of genes simultaneously. The analysis of these large.scale gene expression data has become very useful for investigating gene functions and the interactions among the genes. However, there are few data analysis techniques capable of I'ully exploiting this entirely new class of data. ln this paper, an hybrid evolutionary algorithm has been presented for e{ficiently attaining the skeletal structure of the biomolecular networks and estimating the effective regulatory parameters liom the gene expression time.series data using the recurrent neural network formalism. The suitability of the proposed method has been verified in gene nefwork reconstruction experiments. The reconstruction method has suicessfulty inflrred the underlying network topology and the regulatory parameters while maintaining high accuracy. For the purposes of validation, the proposed methodology has been applied for analyzing the real expression data set of SOS DNA repair system in .Esci erichia coli and succeeded to reconstruct the network of key regulators.

1. ref1