Researches On Biological Network Identification And Reconstruction

Facing long-term accumulation of massive biological data,modeling biological networks and system optimization research can further explore the deep-level information contained in biological networks,which is of great significance for disease diagnosis and drug development.With the increase in the scale of biological data,how to combine experimental data generated by biotechnology with information from computer science,control theory,and other fields,and use machine learning methods to promote the identification and reconstruction of biological networks is a current research hotspot and a difficult point.This paper is oriented to biological networks,and conducts the following research on the network parameter estimation,network reconstruction,and cluster analysis1)Due to its wide application range,heuristic algorithms have been widely used because of their low requirements for computing models.Harmony search algorithm is a better method in global optimization algorithm,because it does not depend on the choice of initial value and has the advantage of stronger global search ability compared with other heuristic algorithms.It is gradually used in the parameter estimation of biological networks.attract attention.In order to find the optimal input signal of the ODE model under the D-Optimal principle,this paper proposes a harmony search algorithm with an elite strategy,and combines the designed optimal input signal to obtain the optimal parameters of the biological network model.The experimental results of a signal transduction network with 23 unknown parameters verify the fast convergence speed and high accuracy of the proposed method2)Model integration is currently an important research direction to improve the accuracy of biological network reconstruction.Accurate network reasoning results can find the regulatory relationship between genes,which can greatly assist in the discovery and research of the cause of disease.This paper combines tree-search-based machine learning methods and Pearson correlation coefficient,and proposes a new weighting method to integrate the results of three basic models.The DREAM4 and DREAM5 challenge datasets as well as the E.coli dataset have shown good inference accuracy,and have been verified on multi-source information fusion.3)Compared with the data obtained by the second-generation sequencing technology,single-cell experimental data can more accurately reflect the status of a single gene rather than the overall average in a statistical sense.It can provide more information on research of the discovery of unknown cell subtypes and cell clustering.To this end,a new clustering algorithm is proposed in this paper.Three distance calculation methods are combined to construct a new distance matrix.At the same time,a small amount of known prior information is used to select the k-means initialization parameter k and the cluster center points.Finally,the proposed method is compared with various clustering methods on a variety of single-cell datasets,and the accuracy and stability of the proposed clustering algorithm are verified.