Abstract 4946: ssTimiGP: Equip the TimiGP system with a single-sample module, facilitating the cell-cell interaction inference of the tumor microenvironment in any sample

Abstract

Abstract Understanding the tumor microenvironment (TME) is essential in cancer treatment and risk assessment. While single-cell RNA-seq (scRNA-seq) offers in-depth insights into the TME, its expensive cost results in limited data sets and size. In contrast, microarray and RNA-seq provide rich datasets from larger cohorts, but they yield bulk data with lower resolution. To leverage the strengths of both approaches and comprehensively study the TME, we developed the computational framework TimiGP (Tumor Immune Microenvironment Illustration through Gene Pairs). It is inspired by the dynamic balance of the immune system and extended to the entire TME. TimiGP utilizes scRNA-seq to define cell type of interest, and bulk transcriptomic profiles to deduce clinically relevant cell-cell interaction networks and cell functions. Our recent release includes both prognosis and response modules to study the association between TME and the corresponding clinical outcomes. However, their applicability is constrained by the necessity for paired clinical information. In response to this limitation, we present a significant enhancement, a single sample module slightly modified from the TimiGP framework (ssTimiGP). This upgraded version enables TME analysis in any individual sample, requiring only bulk RNA-seq data and cell-type markers. It exports a cell-cell interaction network representing the relative expression of functional markers and the abundance of cells. This optimization expands the utility of TimiGP, making it more accessible across diverse samples and species. We have further expanded the application of ssTimiGP from solid tumors to liquid samples (e.g., peripheral blood). To validate its performance, we conducted a comparative analysis between the interactions of eight immune cell types observed in experimental results using FACS and the computational results derived from ssTimiGP based on paired RNA-seq data. As a result, ssTimiGP not only successfully identified interactions between different cell types that exhibited a significant correlation with experimental findings (e.g., CD4+ naïve T cells → Monocyte, R=0.75, p< 0.001, accuracy = 0.8), but also demonstrated its capability to achieve higher resolution in capturing interactions between immune cell subpopulations (e.g., CD4 naïve T cells → activated CD4+ memory T cells, R=0.57, p=0.008, accuracy = 1.0). In our benchmark analysis, ssTimiGP demonstrated superior performance compared to CIBERSORTx (e.g., CD4+ naïve T cells → Monocyte, R=0, Accuracy=0.3). In summary, ssTimiGP signifies an expansion of the TimiGP system, enabling the inference of the TME in individual solid tumors and liquid samples at various resolutions by integrating the strengths of both scRNA-seq and bulk RNA-seq. This approach offers substantial potential for advancing personalized cancer treatment. Citation Format: Chenyang Li, Jianjun Zhang, Chao Cheng. ssTimiGP: Equip the TimiGP system with a single-sample module, facilitating the cell-cell interaction inference of the tumor microenvironment in any sample [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4946.