Taif M. Models and methods of reverse engineering of genetic regulatory networks based on hybrid immune systems

The object of the study is nonlinear nonstationary processes in gene regulatory networks. The purpose of the study is to increase the efficiency of the dynamic gene regulatory network identification process through the new model's development, methods for solving the problem of building hybrid immune systems. Research methods are probability theory methods and mathematical statistics methods, graph theory methods, evolutionary algorithms, systems of nonlinear differential equations, neural network theory, wavelet theory, methods of design, and implementation of information systems for research automation. For the first time, a hybrid method of reconstruction of the right part of the system of ordinary differential equations was developed, which is used to describe the gene regulatory network dynamics, where a wavelet neural network is studied for the computational model; transformation method, which allows to consistently transform the solution space in the reconstruction of gene regulatory networks based on the data of time series of gene expression profiles to search for relationships between the components of the timing. Further development was received: the procedure of the gene regulatory network reconstruction due to the use of the clonal selection algorithm and trigonometric differential evolution; inductive methods of density clustering by combining and using two-step density algorithms, which allows you to remove non-informative genes when processing data from DNA microchips. Methods of clonal selection algorithms hybridization and differential evolution in the gene regulatory networks reconstruction have been improved. The proposed approach allows increasing the convergence and accuracy of the optimization algorithm in solving the problem of the S-system identification.