Описан гибридный метод реконструкции генной регуляторной сети по временным рядам данных экспрессии генов. В качестве модели сети предложена система обыкновенных дифференциальных уравнений. Для проверки эффективности метода проведены исследования на двух моделях искусственных регуляторных сетей.
Розроблено гібридний метод реконструкції генної регуляторної мережі у часових рядах даних експресії генів. Як модель запропоновано систему звичайних диференціальних рівнянь. Для перевірки ефективності методу проведено дослідження на двох моделях штучних регуляторних мереж.
Introduction. Although there are a variety of models and methods for gene regulatory networks reconstruction, the problem of obtaining an adequate model based on experimental data is still urgent. In this regard, many studies use a fixed record of the differential equations based on the S-system. A significant disadvantage of such fixed record of the differential equations is the lack of flexibility of the model, what limits the scope of its application. Purpose. The purpose of this work is development of the hybrid procedure of the solution of the gene networks reconstruction problem based on the ordinary differential equations. Method. Models of the ordinary differential equations are used to model the gene regulatory networks. To solve the differential equations, wavelet-neural networks are used. The topology and tuning of the parameters is determined using the algorithm of the clonal selection. To find the concentration of gene expression products, which are represented by the method of solving the Cauchy problem, the Runge-Kutta method of the fourth order is applied. Results. A hybrid method is developed that implemented the procedure for reconstructing gene regulatory networks based on the gene expression data. The effectiveness of the proposed method is proved by experimental studies that confirm the applicability of this approach to find the relationships between the components of the GRN. Conclusion. The proposed work is a new Wavelet Neural Network and Clonal Algorithm approach for inferring Gene Regulatory Network which is expressed in terms of the ordinary differential model. The result of the proposed procedure is that further improvement of the technology, combined with preprocessing methods, will allow the effective reconstruction of real GRN. The main directions of further research we have chosen to create a meta-procedure for automatic configuration of parameters of a hybrid algorithm. This meta-procedure will reduce the search space by dynamically changing the intervals of representation of the elements values that make up the individuals of the AIS. Using this new method the bioinformatics and biologists can infer any Gene Regulatory Network of their interest. Also, they can understand the regulatory mechanism of the specific genes which causes the combat diseases.
