Development of a Machine Learning-Based Autophagy-Related lncRNA Signature to Improve Prognosis Prediction in Osteosarcoma Patients

Guang-Zhi Zhang; Zuo-Long Wu; Chun-Ying Li; En-Hui Ren; Wen-Hua Yuan; Ya-Jun Deng; Qi-Qi Xie

doi:10.3389/fmolb.2021.615084

Development of a Machine Learning-Based Autophagy-Related lncRNA Signature to Improve Prognosis Prediction in Osteosarcoma Patients

Front Mol Biosci. 2021 May 21:8:615084. doi: 10.3389/fmolb.2021.615084. eCollection 2021.

Authors

Guang-Zhi Zhang^{1

2

3}, Zuo-Long Wu^{1

2}, Chun-Ying Li⁴, En-Hui Ren^{1

2

5}, Wen-Hua Yuan^{1

2}, Ya-Jun Deng^{1

2}, Qi-Qi Xie^{6

7

8}

Affiliations

¹ The Second Clinical Medical College, Lanzhou University, Lanzhou, China.
² Department of Orthopedics, Lanzhou University Second Hospital, Lanzhou, China.
³ Lintao County Traditional Chinese Medicine Hospital of Gansu Province, Lintao, China.
⁴ The Fourth People's Hospital of Qinghai Province, Xining, China.
⁵ Department of Orthopaedics, Xining First People's Hospital, Xining, China.
⁶ Affiliated Hospital of Qinghai University, Xining, China.
⁷ Affiliated Cancer Hospital of Qinghai University, Xining, China.
⁸ Breast Disease Diagnosis and Treatment Center, Affiliated Hospital of Qinghai University & Affiliated Cancer Hospital of Qinghai University, Xining, China.

Abstract

Background: Osteosarcoma is a frequent bone malignancy in children and young adults. Despite the availability of some prognostic biomarkers, most of them fail to accurately predict prognosis in osteosarcoma patients. In this study, we used bioinformatics tools and machine learning algorithms to establish an autophagy-related long non-coding RNA (lncRNA) signature to predict the prognosis of osteosarcoma patients.

Methods: We obtained expression and clinical data from osteosarcoma patients in the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) and Gene Expression Omnibus (GEO) databases. We acquired an autophagy gene list from the Human Autophagy Database (HADb) and identified autophagy-related lncRNAs by co-expression analyses. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses of the autophagy-related lncRNAs were conducted. Univariate and multivariate Cox regression analyses were performed to assess the prognostic value of the autophagy-related lncRNA signature and validate the relationship between the signature and osteosarcoma patient survival in an independent cohort. We also investigated the relationship between the signature and immune cell infiltration.

Results: We initially identified 69 autophagy-related lncRNAs, 13 of which were significant predictors of overall survival in osteosarcoma patients. Kaplan-Meier analyses revealed that the 13 autophagy-related lncRNAs could stratify patients based on their outcomes. Receiver operating characteristic curve analyses confirmed the superior prognostic value of the lncRNA signature compared to clinically used prognostic biomarkers. Importantly, the autophagy-related lncRNA signature predicted patient prognosis independently of clinicopathological characteristics. Furthermore, we found that the expression levels of the autophagy-related lncRNA signature were significantly associated with the infiltration levels of different immune cell subsets, including T cells, NK cells, and dendritic cells.

Conclusion: The autophagy-related lncRNA signature established here is an independent and robust predictor of osteosarcoma patient survival. Our findings also suggest that the expression of these 13 autophagy-related lncRNAs may promote osteosarcoma progression by regulating immune cell infiltration in the tumor microenvironment.

Keywords: autophagy-related lncRNA; immune cell infiltration; osteosarcoma; prognostic signature; survival.