Identification Of Cell Types Based On Single-cell Sequence Data

With the development of single-cell sequencing technology,there will be more single-cell expression data.By applying of single cell gene expression,we can better understand the organization's cellular heterogeneity and potential cell mechanism.However,the complexity of the data brings us many challenges.Cell identity and function can be characterized at the molecular level by unique transcriptomic sig-natures.The general method is to cluster cells that belong to the same cell types based on gene expression data.There have a biological hypothesis that the level of gene expression determines the synthesis of its functional proteins.So similar functional proteins have been synthesized by similar levels of gene expression.Despite some previous work have been undertaken to cluster single-cell data.such as,pcaReduce which is a hierarchical clustering model based on dimension re-duction,t-SNE which is non-linear dimension reduction approach,ZIFA which con-sider dropout rate as a function,RaceID which can handle rare cell types,network clustering named snn-clique model etc.Alought that method have solved clustering question in many view.The filter gene procedures that have been applied to data are rarely taking into account,Which can expand dimension and have bad influences on clustering.To solve those problem our particular novel snn adjacent table and cohesiveness induced by clusterone model.The method called snn-clusterone is able to automatically determine the number of clusters in the data.Moreover,it utilze robust and stable performance of snn enough.Additionally,it avoid the weakness of snn-clique model that ignore overall connectivity of network.The simulation ex-periment have prove that the robust of our model and the better clustering result than snn-clique.From real data,the same result is repeted again in many distance situation.