Identification of molecular subtypes for integrated multi-omics analysis for use in guiding precision medicine in hepatocellular carcinoma.

2576 Background: Tumor microenvironment (TME) heterogeneity leads to the discrepancy of survival prognosis and clinical treatment response for HCC patients. Although the documented molecular subtypes moderately describe the TME heterogeneity and characteristics, their clinical application is constrained by several issues. Methods: We integrated three single-cell datasets from 39 HCC patients to describe the TME landscape and identified prognosis-related cell subclusters. Unsupervised clustering of subcluster-specific markers (SSMs) was performed to generate transcriptomic subtypes. The predictive value of these molecular subtypes for prognosis and treatment response was explored in multiple external HCC cohorts and Xiangya real-world HCC cohort with patients who received immune checkpoint blockade (ICB) therapy. Cancer stemness was estimated using bioinformatic methods and in vivo and in vitro experiments. The features of TME were further validated using single-cell RNA-seq and immune repertoire sequencing, mass cytometry (CyTOF) and multiplex immunofluorescence (mIF). Ultimately, we constructed prognosis-related score (PRS)based on machine learning algorithm, and identified the potential therapeutic targets and agents for high-PRS patients. Results: The comprehensive single-cell analysis provided a high-resolution depiction of TME heterogeneity in HCC and confirmed 6 cell subclusters of prognostic relevance. Five transcriptomic subtypes were constructed using SSMs, which possessed different clinical prognosis, stemness characteristics, immune landscape and therapeutic responses. Class 1 indicated an inflamed phenotype with better clinical outcomes, while Class 2 and class 4 demonstrated immune-deserted phenotypes lacking T cells infiltration. Class 5 and class 3 indicated inhibitory tumor immune microenvironment enriched regulatory T cells and suppressive immune checkpoints. Multiple therapeutic cohorts suggested that Class 5 and class 3 were sensitive to ICB and targeted therapy, while Class 1 and Class 2 were more responsive to transcatheter arterial chemoembolization treatment (TACE). Class 4 displayed resistant to all conventional HCC therapies. The PRS performed well in prognostic prediction in multiple HCC cohorts. Four potential therapeutic agents and four targets were further identified for high-PRS HCC patients. Conclusions: Our study generated a clinically valid molecular classification, thereby providing guidance for precision medicine in HCC patients.