KL2 Scholar Scope - Characterizing the Pan-Cancer Endogenous Retroviruses Landscapes at Single-Cell Resolution

Abstract: HERVs-derived biomarkers and neoantigens have gained tremendous interests in recent years. Specific HERV signatures have been developed in many cancer types showing strong association with clinical outcomes and immunotherapy response. HERVs also contribute to the adaptive immune response through the production of HERV-derived T cell epitopes on the surface of tumor cells to enhance the visibility of tumor cells to immune surveillance. However, the current research paradigm in HERV studies, which is centered on the “bulk-tissue” RNA sequencing, has significant limitations. Tumors are intricate ecosystems, comprising cancer cells, immune cells, stromal cells, and other types. The conventional method of ''bulk-tissue'' RNA sequencing offers only averaged HERV profiles of all cells combined, masking underlying differences in the cell-type-specific transcriptome. Consequently, the true HERV expression signals driving the tumorigenesis or therapeutic resistance from a specific cell population or cell type can be obscured by the low-resolution bulk-RNAseq data, underscoring the need for more precise methodologies to unravel the complexities of HERV expression within diverse cell types in TME. Single- cell techniques have profoundly impacted cancer studies by providing unprecedented resolution to study tumor transcriptome at single-cell resolution. However, HERV expression profiles are still largely uncharted at single cell level due to the absence of dedicated bioinformatics tools for HERV quantification. In this project, we will develop one of the pioneering bioinformatics tools for HERV quantification at the single-cell level to enable comprehensive HERV analysis in large-scale Pan-Cancer single-cell studies. HERV-based clustering analyses will be performed and compared with mRNA-based cell clusters to explore the correlation between HERV expression and protein-coding gene expression. The associations between HERVs and various biological factors and clinical variables will be examined to suggest important HERV signatures for biomarker development. Machine learning models based on HERV profiles will be constructed to predict clinical outcomes and treatment responses.