Research Of Causality In Cellular Molecules Interaction Network Based On Functional Causal Models

The biochemical mechanism in biological cytology originates from the interaction of internal molecules,and the nonlinear feedback in the biochemical network regulates the internal activities of the cell.With the development of high-throughput technologies in the past two decades,such as large-scale cell detection or single-cell sequencing,allows multivariate detection of a large number of cells under too many experimental conditions and has advanced technical reliability.This makes it easier to measure thousands of molecules in biological cells,but it also presents a more serious challenge in understanding the interactions within cells at the cellular and molecular level.That is,how to extract knowledge from these large-scale cellular datasets and build an interactive network that can represent the causal relationships among them,advance the understanding of basic science,and provide rational explanations for a variety of dominant biological phenomena.In basic biological research,the most common way to study causality is to conduct a Randomized Controlled Trials,but the trials is demanding,time-consuming,expensive,and limited by various ethics.In order to get rid of the various limitations of the Randomized Controlled Trials,More and more causal discovery methods based on observational data and using its statistical characteristics are proposed.However,the traditional causal discovery methods have more or less limitations in actual application scenarios.This paper takes the causal discovery in cell observation data as the research focus,and proposes a cellular molecules interaction network discovery method based on a functional causal model.This method combines the similarity of samples and uses the feature selection method to construct a causal skeleton.It takes conditional independence and distribution symmetry to infer the causal direction between cell molecules,and finally generate a cell molecular interaction network,which clearly shows the causal relationship between molecules with a variable number of nodes and directions.The main work of this paper is as follows:Firstly,the development of the interaction network between molecules in cells is described in detail,and the current research status in causal discovery is analyzed.On this basis,the research focus and significance of this paper are expounded.Secondly,this paper introduces the relevant theory of Bayesian network structure learning and causal structure equation.Aiming at the problem of traditional interaction network discovery,this paper proposes a cell-molecular interaction network discovery method based on functional causality model.Firstly,a method is used to take representative sampling of the observed data,and then a Lasso feature selection method is used to obtain the causal skeleton.Finally,the generative network is used for functional causal modeling to minimize the maximum average deviation between the generated data and the observed data,and according to the symmetry of the joint distribution,the causal direction between the variables is determined.Next,analyze the prediction effect of the model on the public biological cell datasets,compared it with different models and analyze the comparison results.Finally,based on the algorithm proposed in the paper,a verification system for causal discovery of cellular molecules interaction network is designed.Firstly,a detailed requirements analysis is carried out and then a general system framework is constructed.Finally,the main functional modules of the system are designed and implemented in detail.