Machine learning-based prognostic modeling of lysosome-related genes for predicting prognosis and immune status of patients with hepatocellular carcinoma

Wenhua Li; Qianwen Wang; Junxia Lu; Bin Zhao; Yuqing Geng; Xiangwei Wu; Xueling Chen

doi:10.3389/fimmu.2023.1169256

Machine learning-based prognostic modeling of lysosome-related genes for predicting prognosis and immune status of patients with hepatocellular carcinoma

Front Immunol. 2023 May 19:14:1169256. doi: 10.3389/fimmu.2023.1169256. eCollection 2023.

Authors

Wenhua Li^{1

2}, Qianwen Wang^{1

2}, Junxia Lu^{1

2}, Bin Zhao^{1

2}, Yuqing Geng^{1

2}, Xiangwei Wu^{1

2

3}, Xueling Chen^{1

2}

Affiliations

¹ Key Laboratory for Prevention and Treatment of High Morbidity in Central Asia, National Health and Health Commission, Shihezi, China.
² Department of Immunology, Shihezi University School of Medicine, Shihezi, China.
³ The First Affiliated Hospital, Shihezi University School of Medicine, Shihezi, China.

Abstract

Background: Hepatocellular carcinoma (HCC) is a leading cause of cancer-related deaths worldwide. Lysosomes are organelles that play an important role in cancer progression by breaking down biomolecules. However, the molecular mechanisms of lysosome-related genes in HCC are not fully understood.

Methods: We downloaded HCC datasets from TCGA and GEO as well as lysosome-related gene sets from AIMGO. After univariate Cox screening of the set of lysosome-associated genes differentially expressed in HCC and normal tissues, risk models were built by machine learning. Model effects were assessed using the concordance index (C-index), Kaplan-Meier (K-M) and receiver operating characteristic curves (ROC). Additionally, we explored the biological function and immune microenvironment between the high- and low-risk groups, and analyzed the response of the high- and low-risk groups to immunotherapy responsiveness and chemotherapeutic agents. Finally, we explored the function of a key gene (RAMP3) at the cellular level.

Results: Univariate Cox yielded 46 differentially and prognostically significant lysosome-related genes, and risk models were constructed using eight genes (RAMP3, GPLD1, FABP5, CD68, CSPG4, SORT1, CSPG5, CSF3R) derived from machine learning. The risk model was a better predictor of clinical outcomes, with the higher risk group having worse clinical outcomes. There were significant differences in biological function, immune microenvironment, and responsiveness to immunotherapy and drug sensitivity between the high and low-risk groups. Finally, we found that RAMP3 inhibited the proliferation, migration, and invasion of HCC cells and correlated with the sensitivity of HCC cells to Idarubicin.

Conclusion: Lysosome-associated gene risk models built by machine learning can effectively predict patient prognosis and offer new prospects for chemotherapy and immunotherapy in HCC. In addition, cellular-level experiments suggest that RAMP3 may be a new target for the treatment of HCC.

Keywords: RAMP3; drug sensitivity; hepatocellular carcinoma; immune infiltration; lysosome; machine learning; prognostic model.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Carcinoma, Hepatocellular* / genetics
Fatty Acid-Binding Proteins
Humans
Immunotherapy
Liver Neoplasms* / genetics
Lysosomes
Prognosis
Tumor Microenvironment / genetics

Substances

FABP5 protein, human
Fatty Acid-Binding Proteins

Grants and funding

This research was supported by Corps guiding science and technology plan project(2022ZD041) and the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences(2020-PT330-003).