Open AccessAn N<sup>6</sup>-methyladenosine and target genes-based study on subtypes and prognosis of lung adenocarcinoma
Author(s) -
Xiao Chu,
Wang Wei-qing,
Zhaoyun Sun,
Feichao Bao,
Feng Li
Publication year - 2022
Publication title -
mathematical biosciences and engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.451
H-Index - 45
eISSN - 1551-0018
pISSN - 1547-1063
DOI - 10.3934/mbe.2022013
Subject(s) - nomogram , proportional hazards model , adenocarcinoma , subtyping , oncology , gene , lung cancer , survival analysis , medicine , biology , computational biology , cancer research , cancer , genetics , computer science , programming language
Purpose: Lung adenocarcinoma (LUAD) is a highly lethal subtype of primary lung cancer with a poor prognosis. N6-methyladenosine (m 6 A), the most predominant form of RNA modification, regulates biological processes and has critical prognostic implications for LUAD. Our study aimed to mine potential target genes of m 6 A regulators to explore their biological significance in subtyping LUAD and predicting survival. Methods: Using gene expression data from TCGA database, candidate target genes of m 6 A were predicted from differentially expressed genes (DEGs) in tumor based on M 6 A2 Target database. The survival-related target DEGs identified by Cox-regression analysis was used for consensus clustering analysis to subtype LUAD. Uni-and multi-variable Cox regression analysis and LASSO Cox-PH regression analysis were used to select the optimal prognostic genes for constructing prognostic score (PS) model. Nomogram encompassing PS score and independent prognostic factors was built to predict 3-year and 5-year survival probability. Results: We obtained 2429 DEGs in tumor tissue, within which, 1267 were predicted to m 6 A target genes. A prognostic m 6 A-DEGs network of 224 survival-related target DEGs was established. We classified LUAD into 2 subtypes, which were significantly different in OS time, clinicopathological characteristics, and fractions of 12 immune cell types. A PS model of five genes (C1QTNF6, THSD1, GRIK2, E2F7 and SLCO1B3) successfully split the training set or an independent GEO dataset into two subgroups with significantly different OS time (p < 0.001, AUC = 0.723; p = 0.017, AUC = 0.705).A nomogram model combining PS status, pathologic stage, and recurrence was built, showing good performance in predicting 3-year and 5-year survival probability (C-index = 0.708, 0.723, p-value = 0). Conclusion: Using candidate m 6 A target genes, we obtained two molecular subtypes and designed a reliable five-gene PS score model for survival prediction in LUAD.