Independent Validation of Early-Stage Non-Small Cell Lung Cancer Prognostic Scores Incorporating Epigenetic and Transcriptional Biomarkers With Gene-Gene Interactions and Main Effects

Abstract

BackgroundDNA methylation and gene expression are promising biomarkers of various cancers, including non-small cell lung cancer (NSCLC). Besides the main effects of biomarkers, the progression of complex diseases is also influenced by gene-gene (G×G) interactions.Research QuestionWould screening the functional capacity of biomarkers on the basis of main effects or interactions, using multiomics data, improve the accuracy of cancer prognosis?Study Design and MethodsBiomarker screening and model validation were used to construct and validate a prognostic prediction model. NSCLC prognosis-associated biomarkers were identified on the basis of either their main effects or interactions with two types of omics data. A prognostic score incorporating epigenetic and transcriptional biomarkers, as well as clinical information, was independently validated.ResultsTwenty-six pairs of biomarkers with G×G interactions and two biomarkers with main effects were significantly associated with NSCLC survival. Compared with a model using clinical information only, the accuracy of the epigenetic and transcriptional biomarker-based prognostic model, measured by area under the receiver operating characteristic curve (AUC), increased by 35.38% (95% CI, 27.09%-42.17%; P = 5.10 × 10–17) and 34.85% (95% CI, 26.33%-41.87%; P = 2.52 × 10–18) for 3- and 5-year survival, respectively, which exhibited a superior predictive ability for NSCLC survival (AUC3 year, 0.88 [95% CI, 0.83-0.93]; and AUC5 year, 0.89 [95% CI, 0.83-0.93]) in an independent Cancer Genome Atlas population. G×G interactions contributed a 65.2% and 91.3% increase in prediction accuracy for 3- and 5-year survival, respectively.InterpretationThe integration of epigenetic and transcriptional biomarkers with main effects and G×G interactions significantly improves the accuracy of prognostic prediction of early-stage NSCLC survival. DNA methylation and gene expression are promising biomarkers of various cancers, including non-small cell lung cancer (NSCLC). Besides the main effects of biomarkers, the progression of complex diseases is also influenced by gene-gene (G×G) interactions. Would screening the functional capacity of biomarkers on the basis of main effects or interactions, using multiomics data, improve the accuracy of cancer prognosis? Biomarker screening and model validation were used to construct and validate a prognostic prediction model. NSCLC prognosis-associated biomarkers were identified on the basis of either their main effects or interactions with two types of omics data. A prognostic score incorporating epigenetic and transcriptional biomarkers, as well as clinical information, was independently validated. Twenty-six pairs of biomarkers with G×G interactions and two biomarkers with main effects were significantly associated with NSCLC survival. Compared with a model using clinical information only, the accuracy of the epigenetic and transcriptional biomarker-based prognostic model, measured by area under the receiver operating characteristic curve (AUC), increased by 35.38% (95% CI, 27.09%-42.17%; P = 5.10 × 10–17) and 34.85% (95% CI, 26.33%-41.87%; P = 2.52 × 10–18) for 3- and 5-year survival, respectively, which exhibited a superior predictive ability for NSCLC survival (AUC3 year, 0.88 [95% CI, 0.83-0.93]; and AUC5 year, 0.89 [95% CI, 0.83-0.93]) in an independent Cancer Genome Atlas population. G×G interactions contributed a 65.2% and 91.3% increase in prediction accuracy for 3- and 5-year survival, respectively. The integration of epigenetic and transcriptional biomarkers with main effects and G×G interactions significantly improves the accuracy of prognostic prediction of early-stage NSCLC survival. Lung cancer is a leading cause of cancer-related death worldwide and was estimated to cause 1.76 million deaths in 2018.1Bray F. Ferlay J. Soerjomataram I. et al.Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.CA Cancer J Clin. 2018; 68: 394-424Crossref PubMed Scopus (33702) Google Scholar The 5-year survival rate among patients with lung cancer remains relatively low, ranging from 4% to 17% depending on clinical characteristics.2Hirsch F.R. Scagliotti G.V. Mulshine J.L. et al.Lung cancer: current therapies and new targeted treatments.Lancet. 2017; 389(10066): 299-311Abstract Full Text Full Text PDF Scopus (1048) Google Scholar Compared with patients diagnosed with late-stage disease, early-stage patients often have a considerably more favorable prognosis. However, significant heterogeneity in clinical prognosis is observed for patients with early-stage non-small cell lung cancer (NSCLC) with similar clinical characteristics, which indicates the importance of understanding molecular mechanisms.3Tang S. Pan Y. Wang Y. et al.Genome-wide association study of survival in early-stage non-small cell lung cancer.Ann Surg Oncol. 2015; 22(2): 630-635Crossref Scopus (41) Google Scholar Identifying molecular changes in oncogene and/or tumor suppressor genes that are associated with NSCLC survival is helpful for developing targeted therapies to prolong patients’ survival time. DNA methylation is a heritable, reversible, epigenetic modification that affects the spatial conformation of DNA and regulates gene expression.4Egger G. Liang G. Aparicio A. et al.Epigenetics in human disease and prospects for epigenetic therapy.Nature. 2004; 429: 457-463Crossref PubMed Scopus (2322) Google Scholar,5Feinberg A.P. Tycko B. The history of cancer epigenetics.Nat Rev Cancer. 2004; 4(2): 143-153Crossref Scopus (1684) Google Scholar DNA methylation is a molecular biomarker and may be a therapeutic target for the treatment of cancer.6Shen S. Zhang R. Guo Y. et al.A multi-omic study reveals BTG2 as a reliable prognostic marker for early-stage non-small cell lung cancer.Mol Oncol. 2018; 12: 913-924Crossref PubMed Scopus (16) Google Scholar,7Wei Y. Liang J. Zhang R. et al.Epigenetic modifications in KDM lysine demethylases associate with survival of early-stage NSCLC.Clin Epigenetics. 2018; 10(1): 41Crossref Scopus (11) Google Scholar In addition, gene-gene (G×G) interactions have long been recognized to regulate the progression of complex diseases, including NSCLC.8Zhang R. Lai L. He J. et al.EGLN2 DNA methylation and expression interact with HIF1A to affect survival of early-stage NSCLC.Epigenetics. 2019; 14: 118-129Crossref PubMed Scopus (10) Google Scholar The development of cancer may be related to interactions between several key genes.9Lin Z. Hui L. Yufei H. et al.Cancer progression prediction using Gene Interaction Regularized Elastic Net.IEEE/ACM Trans Comput Biol Bioinform. 2017; 14: 145-154Crossref PubMed Scopus (19) Google Scholar Lung cancer prognosis-associated biomarkers have been proposed on the basis of omics data, including DNA methylation,10Sandoval J. Mendez-Gonzalez J. Nadal E. et al.A prognostic DNA methylation signature for stage I non-small-cell lung cancer.J Clin Oncol. 2013; 31(32): 4140-4147Crossref Scopus (190) Google Scholar gene expression,11Shedden K. Taylor J.M. Enkemann S.A. et al.Gene expression-based survival prediction in lung adenocarcinoma: a multi-site, blinded validation study.Nat Med. 2008; 14(8): 822-827Google Scholar microRNA,12Tan X. Qin W. Zhang L. et al.A 5-microRNA signature for lung squamous cell carcinoma diagnosis and hsa-miR-31 for prognosis.Clin Cancer Res. 2011; 17(21): 6802-6811Crossref Scopus (159) Google Scholar and long noncoding RNA.13Zhou M. Guo M. He D. et al.A potential signature of eight long non-coding RNAs predicts survival in patients with non-small cell lung cancer.J Transl Med. 2015; 13(1): 231Crossref Scopus (141) Google Scholar However, most studies are limited to a single type of omics data, which results in less accurate prognostic models.14Zhao Q. Shi X. Xie Y. et al.Combining multidimensional genomic measurements for predicting cancer prognosis: observations from TCGA.Brief Bioinform. 2015; 16(2): 291-303Crossref Scopus (85) Google Scholar For example, our previous integrative omics study of the BTG2 gene showed that this gene could slightly improve the prediction accuracy of early-stage NSCLC survival.6Shen S. Zhang R. Guo Y. et al.A multi-omic study reveals BTG2 as a reliable prognostic marker for early-stage non-small cell lung cancer.Mol Oncol. 2018; 12: 913-924Crossref PubMed Scopus (16) Google Scholar However, a large-scale integrative analysis of multiomics data has identified genes with either important main effects or gene-gene (G×G) interactions, based on which a more accurate prognostic prediction model of NSCLC can be constructed. Specifically, we used a two-stage study design and performed an integrative analysis of pan-cancer-related genes to identify prognostic biomarkers with either a main effect or G×G interactions using epigenome and transcriptome data from multiple study centers. We then built a prognostic prediction model for early-stage NSCLC by incorporating both selected epigenetic and transcriptional biomarkers. Only patients with early-stage (stage I or II) lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) were included in our study. DNA methylation data were harmonized from five international study centers, including Harvard, Spain, Norway, Sweden, and the Cancer Genome Atlas (TCGA). Gene expression data were composed of four datasets from the Gene Expression Omnibus (GEO) and TCGA. The Harvard cohort consisted of patients seen at Massachusetts General Hospital (MGH), and histologically confirmed as having primary NSCLC, recruited since 1992.15Guo Y. Zhang R. Shen S. et al.DNA Methylation of LRRC3B: a biomarker for survival of early-stage non-small cell lung cancer patients.Cancer Epidemiol Biomarkers Prev. 2018; 27: 1527-1535Crossref PubMed Scopus (6) Google Scholar We profiled 151 early-stage patients from this cohort. A lung pathologist at MGH evaluated each specimen for the amount (tumor cellularity, > 70%) and quality of tumor cells. The specimens were classified histologically according to World Health Organization criteria. The institutional at the Harvard H. of Health and MGH the study. patients The cohort included patients with early-stage NSCLC recruited from eight between and J. Mendez-Gonzalez J. Nadal E. et al.A prognostic DNA methylation signature for stage I non-small-cell lung cancer.J Clin Oncol. 2013; 31(32): 4140-4147Crossref Scopus (190) Google Scholar and were study was by the institutional The cohort consisted of patients with early-stage NSCLC from recruited between and et al.Genome-wide DNA methylation in lung association with and gene expression and Oncol. PubMed Scopus Google Scholar The was with the of the and patients were in and at DNA DNA was from patients with early-stage NSCLC, including patients with and patients with at the Hospital in A. M. M. et al.Genome-wide DNA methylation analysis of lung carcinoma reveals and four adenocarcinoma associated with Cancer Res. Scopus Google Scholar The study was under the of the in and A of and with DNA survival and data were DNA methylation data from patients with early-stage NSCLC were on information from patients with early-stage NSCLC was profiled using the Genome Only data from patients with survival clinical and tumor expression were DNA methylation was with data were Methylation to methylation and to and quality were of the quality > in of were or were in or in M. S. et of and in the 2013; Scopus Google or (6) data in centers. with > were Methylation were for in as well as type I and in were for effects in according to the by a F. M. et of analysis for DNA methylation using the 2013; PubMed Scopus Google Scholar of the are in The the data and were using by gene data were from the data and were for Gene were the rate > and the effect was with The expression of each gene was on a and association DNA methylation and gene expression of pan-cancer-related genes were then used for association Gene for the pan-cancer-related genes were from the of in Cancer were identified for association from five genes S. Zhang R. Guo Y. et al.A multi-omic study reveals BTG2 as a reliable prognostic marker for early-stage non-small cell lung cancer.Mol Oncol. 2018; 12: 913-924Crossref PubMed Scopus (16) Google Scholar Y. Liang J. Zhang R. et al.Epigenetic modifications in KDM lysine demethylases associate with survival of early-stage NSCLC.Clin Epigenetics. 2018; 10(1): 41Crossref Scopus (11) Google Scholar R. Lai L. He J. et al.EGLN2 DNA methylation and expression interact with HIF1A to affect survival of early-stage NSCLC.Epigenetics. 2019; 14: 118-129Crossref PubMed Scopus (10) Google Scholar Y. Zhang R. Shen S. et al.DNA Methylation of LRRC3B: a biomarker for survival of early-stage non-small cell lung cancer patients.Cancer Epidemiol Biomarkers Prev. 2018; 27: 1527-1535Crossref PubMed Scopus (6) Google Scholar and R. Lai L. X. et methylation the of on lung adenocarcinoma an Oncol. 2019; Scopus Google in our previous study were also The of analysis is in and transcriptional were performed and a and validation were used to identify NSCLC prognostic biomarkers. In each we analysis of both the main effects and gene-gene interactions among biomarkers. having DNA methylation data from Harvard, Spain, Norway, and Sweden, as well as patients having gene expression data from were to the for epigenetic analysis and transcriptional respectively. having two types of omics data from were to the validation For the main effect we used screening and to biomarkers with main effects that were to survival, using the is a two-stage the the biomarkers with the with survival. the was used to and among the biomarkers selected at the the was based on information criteria. biomarkers that be at the we the to the biomarkers new biomarkers can be Y. an for screening in 2018; Scholar is for the heterogeneity between and we used a model to biomarkers, using the survival. The model for the between and in The in the model were study clinical and For the G×G a model for the was to identify biomarkers with G×G interactions. The P for multiple were by the which the to by the of the type I be at the In our the of G×G analysis of epigenetic and transcriptional biomarkers was as × = × and × = × respectively. biomarkers observed in the were confirmed in the validation and were the P was and was of the effect two We also performed a of for each significant The and were as of DNA methylation or gene expression analysis was performed to significant biomarkers. were their DNA methylation or expression were of based on × For identified biomarkers, we a to a model in the which was then in In the a was to the main effect or G×G of biomarkers and > analysis was also performed using two and and transcriptional were on the basis of a of of the DNA methylation and gene with from the model. were by epigenetic and transcriptional the prognostic score was as the of clinical information and integrative score survival for the were to the survival among patients with We 3- and 5-year survival of using the for for survival data and a Scopus Google Scholar The accuracy of the prediction is using a receiver operating characteristic curve and was measured by area under the curve (AUC), by the The prediction accuracy was confirmed with an independent in the validation The and P of the were on the basis of analysis of prognostic was by clinical and The an accuracy of predictive survival as well as the CI, which from to were to the predictive J. of the for of survival Res. 2018; Scopus Google Scholar A was with to of our model. We the potential of the identified genes at the by of limited we evaluated the association between expression and gene using the from the we performed expression analysis between tumor and and the main effects of genes and G×G interactions between genes on survival, using the expression analysis was performed with the which a model to changes and to B. D. et expression for and Res. 2015; Scopus Google Scholar Gene and of and analysis was with Gene analysis was with D. et prediction integration for gene and predicting gene Res. Scopus Google Scholar a of the The to in a were as the P were were performed with = and = patients with and gene were included in the association The and clinical information are in For the main effect analysis of DNA methylation and gene and gene were selected by respectively. However, was significantly associated with survival in both = [95% CI, P = × = [95% CI, P = × and significant in gene in the gene significant in the validation = [95% CI, P = × = [95% CI, P = × and For the G×G we observed and G×G interactions from epigenetic and transcriptional respectively, in the and G×G interactions were in the validation that were also significant in the analysis analysis in the we observed with a main effect and pairs of with G×G interactions in the model which was used to the epigenetic score = [95% CI, P = × gene with a main effect and of gene with a G×G were in the model and used to the transcriptional score = [95% CI, P = × The between survival and each of were independently confirmed in the validation for = [95% CI, P = × transcriptional = [95% CI, P = × integrative = [95% CI, P = × prognostic = [95% CI, P = × the ability of in the validation were and based on the of and prognostic respectively. Compared with the epigenetic the and and mortality = [95% CI, P = × = [95% CI, P = × with a transcriptional score significantly survival = [95% CI, P = × = [95% CI, P = × The significant survival was among patients with integrative = [95% CI, P × = [95% CI, P = × with clinical information, including study clinical and the prognostic score significantly NSCLC survival = [95% CI, P = × = [95% CI, P = × 10–17) The ability of the prognostic score is by patients on the basis of the of the an patients in 3- and 5-year survival as well as survival time. indicates that patients with mortality can be by using our score = [95% CI, P = × = [95% CI, P = × = [95% CI, P = × = [95% CI, P = × The of the prognostic score was confirmed in the analysis by of results from analysis of prognostic with of the prognostic score on non-small cell lung cancer survival in various is by clinical = lung = lung squamous cell for of We then independently the predictive ability of biomarkers. The model with clinical information, as limited prediction ability (AUC3 = AUC5 = However, by biomarkers with either main effects or G×G interactions, the significantly increased by 35.38% (95% CI, P = 5.10 × 10–17) and 34.85% (95% CI, 26.33%-41.87%; P = 2.52 × 10–18) for 3- and 5-year survival, respectively, and exhibited a superior predictive ability for NSCLC survival (AUC3 = 0.88 [95% CI, 0.83-0.93]; AUC5 = 0.89 [95% CI, 0.83-0.93]) G×G interactions contributed an 65.2% for and 91.3% for the 5-year prediction accuracy In the we the using two = and and that the of the selected biomarkers were the as in the model We then their with NSCLC survival, and similar results the of our prognostic model using were = = = for = = = for 5-year survival and we that the effects of four significantly between patients with and patients with score = score = score = score = prognostic model similar prediction ability in both the (AUC3 = AUC5 = and = CI, and (AUC3 = AUC5 = and = CI, the of the selected biomarkers and their interactions in predicting the for patients with and patients with of our prognostic prediction model, we clinical information and of biomarkers and a which estimated well a 3- or 5-year survival The of the prognostic score prediction accuracy = CI, in an independent population. The also showed between observed and survival In of the four genes in significant between gene expression and expression of the genes in were between tumor and with In addition, gene with main effect and four pairs of genes with G×G a significant effect on survival genes identified in epigenetic five genes and were identified as genes with the genes were in cancer-related including the non-small cell lung cancer accurate prognostic predictive model may in clinical or for the patients with mortality and tumor have been used as increased has that molecular biomarkers may is tumor may the tumor is and of biomarkers Z. J. Z. et for in 2019; Scopus Google Scholar is a prognostic predictive model both and effects can be using that are both and We a two-stage integrative study of DNA methylation and gene expression data from multiple to a prognostic incorporating biomarkers with main effects and G×G interactions. The prognostic which was in an independent survival for patients with early-stage NSCLC and significantly improves prediction accuracy for their prognosis. G×G interactions are of important the of complex gene-gene interactions that human Rev Scopus Google Scholar was in previous studies that of G×G interactions improve the predictive accuracy of et and targeted of J Med. 2008; Scopus Google Q. M. J. W. of genomic for to diseases and Med. 2004; PubMed Scopus Google Scholar However, interactions improve prediction their effects are or are significant interactions, H. J. of gene-gene and interactions to improve prediction for complex J Full Text Full Text PDF PubMed Scopus Google Scholar Besides G×G interactions could increase the to and then be for the of new gene-gene interactions that human Rev Scopus Google Scholar results showed that biomarkers with G×G interactions significantly and the prognostic prediction accuracy of early-stage NSCLC, which be to increased the prediction accuracy of our model, we a to our studies with The of the screening are in The prediction accuracy of our model is superior as the study with the a study with the independent prediction capacity = study has a relatively and a prediction model that performed well in an independent of (AUC3 = 0.88 and AUC5 = and the genes identified in transcriptional K. A. S. et may tumor and in patients with lung cancer.J Oncol. Full Text Full Text PDF PubMed Scopus Google Scholar and G. G. F. et is a for the of lung 2013; Scopus Google Scholar have been to be associated with lung In this among genes identified in epigenetic five and were identified as genes in the gene is associated with NSCLC L. et al.Genome-wide of that are to non-small cell lung 2018; PubMed Scopus Google Scholar NSCLC cell survival and is a therapeutic target for NSCLC S. J.M. G. et is in non-small-cell lung cancer and associated with J Cancer. 2004; Scopus Google Scholar of the by Y. of the gene and related genes as of in lung Scopus Google Scholar is in to of S. et in non-small cell lung cancer and Cancer Clin Oncol. Scopus Google D. I. et is by independent of key in non-small cell lung Scopus (85) Google R. Z. Z. et of expression with in lung adenocarcinoma and squamous cell 2011; Google Scholar is a of NSCLC prognosis and related to in NSCLC B. A. et oncogene as a predictive marker of Med. 13(1): Full Text Full Text PDF Scopus Google K. K. M. et of and of in human non-small cell lung cancer D. K. The of in non-small cell lung PubMed Scopus Google Scholar is an independent for predicting NSCLC J. Zhang X. H. et expression of is associated with survival in patients with non-small cell lung J Clin Scopus Google Scholar could cell in lung S. et non-small cell lung cancer (NSCLC) development of PubMed Scopus Google Scholar and a in NSCLC genomic R. H. M. A. F. L. The is a target in non-small cell lung cancer (NSCLC) Res. Scholar In genes were significantly in or that are cancer the identified genes were also in the non-small cell lung cancer The genes and in the were also in this The results functional the identified are potential epigenetic for NSCLC study has as studies on main effects of biomarkers, their G×G interactions that for of complex diseases Y. Zhang R. Shen S. et al.DNA Methylation of LRRC3B: a biomarker for survival of early-stage non-small cell lung cancer patients.Cancer Epidemiol Biomarkers Prev. 2018; 27: 1527-1535Crossref PubMed Scopus (6) Google Scholar most studies on single omics data prognostic J. Mendez-Gonzalez J. Nadal E. et al.A prognostic DNA methylation signature for stage I non-small-cell lung cancer.J Clin Oncol. 2013; 31(32): 4140-4147Crossref Scopus (190) Google K. Taylor J.M. Enkemann S.A. et al.Gene expression-based survival prediction in lung adenocarcinoma: a multi-site, blinded validation study.Nat Med. 2008; 14(8): 822-827Google X. Qin W. Zhang L. et al.A 5-microRNA signature for lung squamous cell carcinoma diagnosis and hsa-miR-31 for prognosis.Clin Cancer Res. 2011; 17(21): 6802-6811Crossref Scopus (159) Google M. Guo M. He D. et al.A potential signature of eight long non-coding RNAs predicts survival in patients with non-small cell lung cancer.J Transl Med. 2015; 13(1): 231Crossref Scopus (141) Google Scholar of and data and both G×G interactions and main A. et al.A score for of in Scopus Google Scholar we built prognostic which could improve prognostic to identify reliable prognostic biomarkers for the prediction of early-stage NSCLC survival, we used criteria. In the main effect biomarkers, with effect a in to in the model. For the G×G we the most to for In addition, significant biomarkers observed in the be in an independent population. However, was that a biomarkers were identified of the limited of gene expression data, which contributed a of accuracy of our we used and to biomarkers with main effect and interactions, respectively, and built used as to were from in estimates of effect are more to the complex association and could improve prediction J. S. M. et prediction for disease with a 2011; Scopus Google A. et for association studies in Scopus Google the prediction accuracy of our prognostic model was in as well as by and the genes we identified as in the non-small cell lung cancer and most of the genes have been to be associated with NSCLC, the of our prognostic biomarkers. We also of our as We on most genes the the of B. et al.Cancer 2013; Scopus Google Scholar or can G×G interactions for data on We pairs of pan-cancer-related limited clinical information was for several that were However, in our a clinical and of biomarkers exhibited which for the rate of survival for is relatively which considerably the However, a two-stage this study showed the of our our prognostic prediction model predicts survival and of patients with mortality which a potential for the of and to each of However, information on biomarkers, which clinical of the history of cancer omics has as improves and the is more and to biomarker et of gene and gene in clinical 2017; Scopus Google Scholar In the in our a to clinical and the to and studies with a large-scale and of are to the results of our association study and the of the genes and their interactions. of analysis in and in gene and the functional The prognostic score incorporating biomarkers with both main effects and G×G interactions significantly improves prognostic prediction accuracy for early-stage NSCLC survival.

Topics & Concepts