Development For A Tissue Specific Immune Cell Deconvolution Model And A Comprehensive Database Of Immune Cell Composition From Different Tissues And Diseases

Biological tissues were consisted of many different cell types,among which tissue immune cells often play an important role in multiple situations including both the physiological and pathological conditions.Enumerating for the abundance of tissue immune cells will be helpful for us to study their roles in different situations.Traditional tissue immune cell quantification methods mainly use the labeling of protein molecules specifically expressed by different immune cells to detect the number of different cell types.With the rapid development of high-throughput transcriptome measurement technology,large amount of transcriptome data about different immune cells was accumulated in public databases.Using these data,multiple tissue immune cell calculation models have been developed and used to predict the composition of different immune cells from tissue transcriptome data.In our previous work,two computational tools have been constructed to estimate the composition of tissue immune cells from the transcriptome of mouse profiled on both DNA microarray and RNA-Seq platform.Among all these computational tools,the training data are developed without considering the tissue origin of immune cells.However,as immune cells are localized in different tissues,they may have distinct expression profiles.Hence,calculation with a general training matrices may decrease the calculation accuracy.To reduce the impact introduced by the bias between different tissues,single cell RNA sequencing(scRNA-Seq)data from different mouse tissues could be used to generate tissue-specific signature matrices that are used as the input of the deconvolution model.In this study,the transcriptome of immune cells in each tissue was first extracted from scRNA-Seq data and used to construct the mean expression matrix of tissue immune cells.Then,after comparing different gene selection strategies,the expression of 162 seq-ImmuCC derived signature genes in tissue immune cell scRNA-Seq data were taken as the tissue specific signature matrices.Finally,the performance of this new generated signature matrices were evaluated in multiple tissues and a modest improvement was observed.In addition to optimize for the calculation performance,we also want to apply these computational models to more areas.As immune cell composition is highly divergent across different tissues and diseases,comprehensive resource of tissue immune cells across different conditions in mouse and human will thus provide great understanding of the immune microenvironment among different diseases.Additionally,there is no available database that can be used to query the number of tissue immune cells under different diseases or tissue conditions.Using these computational tools,large-scale estimation of immune cell composition across different conditions should be possible using transcriptome data collected from public databases.Here,266 tissue types and 706 disease types in humans,as well as 143 tissue types and 61 disease types,and 206 genotypes in mouse had been included in a database we have named ImmuCellDB.In ImmuCellDB,users can search and browse immune cell proportions based on tissues,diseases or genotypes in mouse or humans.Additionally,the variation and correlation of immune cell abundance and gene expression level between different conditions can be compared and viewed in this database.In summary,we first construct a tissue-specific quantitative model using the scRNA-Seq data of different tissues of mice.The development of this tool can help us better understand the gene expression difference of each immune cell across different tissues and propose possible strategies to reduce its impact on model calculations.Then,using the computing tools and the transcriptome data deposited in public databases,a database about tissue immune cell composition and gene expression information covering various tissues,diseases or genotype conditions of humans and mice named ImmuCellDB was constructed.Using this database,we can quickly query and evaluate for the immune cell abundance and gene expression value differences under different conditions,as well as the correlation between gene expression values and immune cell number.Altogether,this study not only provides many valuable explorations for the model construction but also enrich our knowledge for tissue immune constitutions across different conditions.