The relationships between gene regulatory functions and hepatocellular carcinoma (HCC) occurrence and progression are constantly being clarified. However, tumour microenvironment complexity has hindered the classification of the role of genes. A comprehensive analysis to further clarify gene functions could provide additional benefits to HCC patients. In the present study, we combined single-cell sequencing data, Mendelian randomization, and bioinformatics analysis for comprehensive analysis. After the study was completed we found that T cell, dendritic cell (DC), macrophage and monocyte contents and the interaction between immune cells in the HCC microenvironment differed between the microvascular invasion-positive (MVI +) and microvascular invasion-negative (MVI-) groups. Mendelian randomization analysis indicated that causal relationships between several endoplasmic reticulum autophagy (ER-phagy) genes and T cell, DC, macrophage and monocyte contents. Single-cell sequencing data were used to validate the association of these genes with immune cells in the microenvironment. Based on the above results, we preliminarily elucidated the potential role of ER autophagy in the HCC microenvironment. Furthermore, a prognostic model was constructed using these causal association genes, which could accurately predict the prognosis and survival of HCC patients.