The Design And Implementation Of Single-cell RNA-seq Subpopulation Visual Analytics System

Advances in single-cell RNA-sequencing technology(sc RNA-seq)have allowed researchers to study cellular heterogeneity and analyze cell subpopulation at single-cell resolution.Cell subpopulation analysis is a key goal of sc RNA-seq data analysis,which can reveal heterogeneity be-tween cells and extract meaningful biological information.Projection is an important visualization method for cell subpopulation analysis.A sub-population projection with high visual separation can visually present the distribution pattern of cell subpopulations and assist users in analyzing and subdividing cell subpopulations.However,subpopulation projection and cell subpopulation analysis present the following challenges:(1)The sub-population structure of cells is usually multi-level,and the heterogeneity between subpopulations at different levels is different,but the existing di-mensionality reduction methods cannot represent multi-level subpopulation structures,resulting in the generated projection results that are visually dif-ficult to distinguish different cell subpopulations;(2)Analysts need to inte-grate their knowledge into the analysis process through interactive methods to obtain more accurate subpopulation analysis results,but there is a lack of a hierarchical interactive analysis process that can effectively analyze datasets with multi-level subpopulation structures.Therefore,this thesis proposes a visual analytics system for single-cell RNA-seq subpopulation.The system allows analysts to explore,analyze and subdivide cell subpop-ulations hierarchically,identify new cell types and generate subpopulation projections with high visual separation.The main research work of this thesis is as follows:(1)This thesis proposes a hierarchical distance measurement method for subpopulation projection,which provides a multi-resolution distance measurement method,which can mine the relationship between cells from multiple levels and obtain the subpopulation projection result with higher visual separation.(2)This thesis proposes a hierarchical subpopulation analysis process and a cluster evaluation index based on gene differences,which evaluates the subpopulation subdivision and can assist users to perform hierarchical interactive exploration and analysis of cell subpopulations.(3)Based on the above methods and processes,this thesis designs and implements a visual analytics system for single-cell RNA-seq subpopula-tion.The system allows users to incorporate domain knowledge into inter-active analysis pipelines,hierarchically explore,analyze and subdivide cell subpopulations,and generate subpopulation projections with higher visual separation.(4)This thesis completed three case experiments in cooperation with domain experts,including subpopulation pattern validation and gene anal-ysis,hierarchical exploration and correction of subpopulation projections,and knowledge-driven cell subpopulation identification and subdivision.The results of three case experiments verify the effectiveness and practi-cability of the system in this thesis.16 Figures,3 Tables,74 References...