Integrating genetic mutations and expression profiles for survival prediction of lung adenocarcinoma

Abstract

Background Lung adenocarcinoma (LUAD) is a set of heterogeneous diseases with distinct genetic and transcriptomic characteristics. Since the introduction of the 2011 International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society histologic classification, increasing evidence has provided insights into genomic mutations and rearrangements among individual histologic subtypes of LUAD. However, how genotypic and phenotypic features of LUAD are interconnected is not well understood. Methods We obtained the genomic, transcriptomic, and clinical data sets of 488 LUAD patients from The Cancer Genome Atlas database. Advanced statistical models were used to disentangle the interactions between genetic mutations and expression profiles, and to assess the alterations and changes in expression of each histologic subtype. The prognostic impacts of genetic mutations, expression profiles, and clinicopathological features were integrated to predict the outcomes of LUAD patients. Results From our data, one or more genetic mutations correlate with expression levels of 6054/18175 (33.3%) genes and explain 8–40% of observed variability in LUAD. The genetic mutations and expression profiles varied remarkably among the histologic subtypes of LUAD, which helped to explain the different prognostic impact based on subtype classification. Genomic, transcriptomic, and clinical data were all shown to have utility for predicting overall and recurrence‐free survival, with the largest contribution from the transcriptome. Conclusion Our prediction model integrating genetic mutations, expression profiles, and clinicopathological features exhibited superior accuracy over the current tumor node metastasis staging system to prognosticate outcomes of patients with LUAD (overall survival 67% vs. 55%, recurrence‐free survival 57% vs. 49%; P < 0.01).

Development of a Novel Prognostic Score Combining Clinicopathologic Variables, Gene Expression and Mutationprofiles for Lung Adenocarcinoma

“…For example, Chen et al [13] integrated two micro-RNAs, two mRNAs and two DNA methylation sites as prognostic factors associated with OS, and they achieved a more significant risk stratification within pathologically-defined subgroups. Song et.al showed that, by integrating genetic mutations and expression profiles with clinicopathologic variables, the prediction of both OS and RFS showed the highest cross-validation accuracy among all the models in the TCGA-LUAD data [15]. Besides, Dong et al [14] found that by adding DNA methylation and gene expression biomarkers to a model using only clinical data as the input, the AUCs improved by 18.3% and 16.4% in discovery and validation phases for early-stage LUAD patients, respectively.…”

“…For example, Li et al [7] reported gene expression-based models with an average C-index of 0.604 on testing datasets from TCGA-LUAD in predicting overall survival (OS). Other studies using multiple types of input data made statistical inference on the significance of potential individual prognostic factors [13][14][15][16]. Two of these studies [15,16] had shown clear benefit of combining genetic mutations and expression profiles in predicting OS and RFS at cross-validation level.…”

Development of a novel prognostic score combining clinicopathologic variables, gene expression, and mutation profiles for lung adenocarcinoma

“…For example, Li et al [7] reported gene expression-based models with an average C-index of 0.604 on testing datasets from TCGA-LUAD in predicting overall survival (OS). Other studies using multiple types of input data made statistical inference on the significance of potential individual prognostic factors [13][14][15][16]. Two of these studies [15,16] had shown clear benefit of combining genetic mutations and expression profiles in predicting OS and RFS at cross-validation level.…”

“…Other studies using multiple types of input data made statistical inference on the significance of potential individual prognostic factors [13][14][15][16]. Two of these studies [15,16] had shown clear benefit of combining genetic mutations and expression profiles in predicting OS and RFS at cross-validation level. In particular, they inferred that the genotype and expression data made around 5% and 50% relative contributions to explained variance of survival outcomes [16].…”

Evaluation of clinical value and potential mechanism of MTFR2 in lung adenocarcinoma via bioinformatics

“…Non-small cell lung cancer (NSCLC) was one of the most common malignant tumors in the world, and lung adenocarcinoma (LUAD) was one of the common subtypes of NSCLC [1,2]. LUAD patients in early stage got improved long-term prognosis by surgery and neoadjuvant chemotherapy, while due to incomplete excision, primary and secondary drug resistance and other reasons, recurrence and metastasis happened.…”