Single-cell RNA-seq Data Preprocessing Algorithm Based On LOESS Regression Weighting

Traditional RNA sequencing technology research is to study gene function and gene structure from the overall level,and has been widely used in basic research,clinical diagnosis and drug research and development.However,the traditional RNAseq method is generally an analysis method for population cells,and the actual result is the average value of the gene expression of a group of cells.Therefore,it is difficult for conventional transcriptome analysis methods to identify various cell subpopulations in heterogeneous populations.Transcriptome characteristics.Singlecell RNA-seq(sc RNA-seq,single-cell transcriptome)technology uses optimized next-generation sequencing technology to obtain transcriptomics information from individual cells to better understand cell functions at the cellular level.But with it,single-cell RNA-seq usually exhibits a higher level of noise and more zeros than RNA-seq data from a large number of cell populations,So how to analyze single-cell RNA-seq data is a very serious challenge.Single-cell RNA-seq data provides us with the opportunity to study cell heterogeneity and differentially expressed genes under biological conditions.Some of the highly mutated genes that have significant changes in the expression of cells have downstream analysis of single-cell sequencing data.The key role,this paper proposes a single-cell RNA-Seq data preprocessing algorithm based on LOESS regression weighting to process the gene expression data in the cell,so that the role of high-variation genes in the entire analysis process is strengthened to achieve gene soft screening and data The purpose of noise reduction.Further,I selected 6 sets of single-cell RNA-seq data to test the algorithm,first preprocess the gene expression matrix generated,and then analyze the impact of pretreatment on subsequent analysis(visualization,clustering,differential expression analysis),experimental results It shows that the algorithm effectively improves the accuracy of downstream analysis and shows good application value.After that,we analyze the impact of the currently used single-cell RNA-seq dimensionality reduction and feature screening algorithms on subsequent analysis.Experiments show that this method has a better effect on subsequent analysis than other methods.Here,we use the new pre-processing method to process the single-cell data of 768 CAFs isolated from breast cancer genetic engineering,and then perform unsupervised clustering,differentially expressed genes,and differentiation trajectory analysis on them.Finally,we analyze Three cell subclasses were found,two cell differentiation nodes were found using the monocle tool,cells were divided into five differentiated states,and genes whose expression levels changed significantly before and after these differentiation nodes were found.The micro-environment has a further understanding.