Identifying Associated Relationships Between Omics Data Via Laplacian Regularized Sparse Matching Model

The rapid development of biological science and technology promotes the emergence of high-throughput sequencing technology,which greatly reduces the cost of sequencing and significantly improves the performance of sequencing.It makes it possible to comprehensively and efficiently conduct sequencing on the same cohort of samples for omics data with different aspects.Vast amounts of multi-source heterogeneous biomedical data have been accumulated,which establish the data foundation for integrative analysis of omics data and provide an unprecedented opportunity for revealing the mechanism of the development and progression of tumors.But due to some factors such as the large noise component in the omics data and the difference between individuals,it is still an unsolved problem to efficiently and accurately identify associated relationships between omics data.Although existing methods have made good progress in identifying associated relationships between omics data,there still exist some disadvantages.For instance,the majority of existing researches only focus on the omics data,and rarely sufficiently utilize the important pre-existing information,or,some of them have not used it at all.However,some studies have indicated that the rational use of pre-existing information can improve the accuracy,robustness,and efficiency of the model.Besides,a growing number of relationships between the biological molecules have been confirmed,such as gene regulatory network,protein-protein interaction networks,and metabolic networks.Based on this,the main contents of this article are as follows:Firstly,we briefly introduced the background,present studying situation,and meaning of identifying associated relationships between omics data.Secondly,we gave a brief introduction to the methods of omics data acquisition and preprocessing.Thirdly,we briefly described the characteristics of Laplacian matrix.Moreover,we put forward the Laplacian regularized sparse matching model?LRSM?for identifying associated relationships between omics data according to their characteristics.To analyze and dig out the potential relationships behind the omics data,Laplacian regularization terms were incorporated into the model.In addition,the model introduced L₀-norm to constraint the sparsity of the result.It can be finally transformed into a convex quadratic optimization problem with inequality constraints and efficiently solved by combining augmented Lagrangian multiplier method and stochastic gradient descent method.Finally,the Laplacian regularized sparse matching model was demonstrated effective and reliable when compared with two benchmark methods through extensive experiments on both synthetic and empirical data.In conclusion,the main significance of this paper lies in:?1?Prior information was introduced by adding Laplacian regularization terms into the model,in order to decrease the impact of noise to the result,reduce the uncertainty caused by data error and improve the accuracy and robustness of the model.?2?Simulation study about the impact of Laplacian regularization terms on the model was conducted to illustrate the important influences of pre-existing information on identifying associated relationships between omics data.?3?In addition to the theoretical analysis,complete biological verification was also conducted to demonstrate the validity and rationality of our model,which has certain positive significance to help deepen the understanding of the regular pattern in tumor genesis and development across multiple omics layers.