The Algorithm Development For Immune Cell Abundance Prediction And Its Application On Tumor Immunotherapy Research

The immune system,comprising various proteins,immune cells,and tissues,is complex and important for host defense.Immune cells in the immune system includes innate and acquired immune cells.Besides,immune cells have many subsets with specific function,thus,investigating immune cell distribution in individuals could provide important insights into immune status,disease progression,prognosis and therapy(particularly in cancer immunotherapy).Therefore,the quantitative analysis of immune cells is very important for cancer progression and treatment related research.In this paper,focusing on the immune cells,we developed immune cell abundance prediction tools and applied them in the tumor immunotherapy research and have achieved the following three research results:1.Immu Cell AI: a tool for comprehensive T-cell subsets abundance predictionThe distribution and abundance of immune cells,particularly T-cell subsets,play pivotal roles in cancer immunology and therapy.T cells has many subsets with specific function and current methods are limited in estimating them,thus,a method for predicting comprehensive T-cell subsets is urgently needed.The immune cell abundance identifier(Immu Cell AI)was developed based on the collecting immune cell expression profiles and signature gene sets by gene enrichment analysis,for precisely estimating the abundance of24 immune cell types including 18 T-cell subsets,from gene expression data.The deviation expression profile was introduced in the algorithm to balance the expression among different immune cell signature gene sets.For each sample,Immu Cell AI first calculates the expression deviation of the sample relative to the immune cell reference profiles,then ss GSEA method is used to calculate the enrichment score of each immune cell signature gene sets in the deviation profile,and the immune cell abudnacne of the sample is the corrected and normalized enrichment score obtained by the last step.Performance evaluation on data with flow cytometry results and expression data indicated that Immu Cell AI can estimate the abundance of immune cells with superior accuracy than other methods especially on many T-cell subsets.To better utilize Immu Cell AI,we designed a user-friendly web server(http://bioinfo.life.hust.edu.cn/Immu Cell AI/),in which users can estimate the abundance of 24 immune cell types,compare the immune cell infiltration difference between samples in different groups,and predict the response to immune checkpoint blockade therapy from gene expression profiles.2.The application of Immu Cell AI in tumor immunotherapy researchImmune checkpoint blockade(ICB)therapy has revolutionized cancer treatment by restarting the anti-tumor immune response by acting directly on the tumor immune microenvironment.The distribution and status of immune cells in tumor tissue play an important role in the immunotherapy process.However,there is a lack of immunotherapy response prediction models characterized by immune cell abundance.In this study,for the first time,an immunotherapy response prediction model was developed using immune cell abundance as feature.A retrospective analysis of published melanoma samples using Immu Cell AI revealed that after treatment,the abundance of dendritic cells and cytotoxic T cells were increased and effector memory T cells decreased,which indicated that immune response was activated in cancer tissues after treatment.By comparing the infiltration between samples in the response and non-response groups,it was found that the abundance of cells with immune killing or helper functions was slightly higher in responders than non-responders before treatment,and the difference between the two groups was more significant after treatment,indicating that patients in the response group were more sensitive to ICB treatment.Moreover,Immu Cell AI result-based model for predicting the ICB response was build with high accuracy(more than 80%)in both the test and the independent validation datasets with different cancer types.3.Immu Cell AI-mouse: a tool for comprehensive mouse immune cell abundance predictionMouse models are widely used in cancer related researches and the quantification of immune cell fraction in mouse TME is very important.Based on the Immu Cell AI,immune cell abundance identifier for mouse(Immu Cell AI-mouse)was developed by employing a cell stratification strategy,which was a tool for estimating the abundance of comprehensive immune cell types and their subtypes from gene expression data(http://bioinfo.life.hust.edu.cn/Immu Cell AI-mouse).First,all 36 cell types are classified into three layers according to the cell lineage of them.Layer 1 is composed of seven major immune cell types: B cells,monocytes,macrophages,T cells,Dendritic cells(DCs),NK cells,and granulocytes.Cells in layer 2 are subtypes of cells in the first layer,such as subtypes of T cells: CD4 T,CD8 T,et al.And cells in layer 3 are subtypes of CD4 T and CD8 T cells.In the prediction process,the abundance of cells in three layers is predicted separately based on the Immu Cell AI method,and then,the result is normalized according to the cell lineage relationships among immune cells.Performance evaluation on bulk samples estimated by RNA-Seq or microarray and simulated samples from single-cell RNA-seq data showed the high accuracy of Immu Cell AI-mouse in predicting most of immune cell types.In conclusion,centered on the immune cells,the immune cell abundance prediction tools(Immu Cell AI and Immu Cell AI-mouse)were developed,which can be used to predict the abundance of immune cells derived from human or mouse samples.Besides,retrospective analysis of samples from ICB-treated patients or mouse models were performed using the above tools,and an Immu Cell AI result-based ICB treatment response prediction model was build,which provides important methods and clues for tumor microenvironment and immunotherapy-related studies.