A comprehensive analysis of metabolomics and transcriptomics in non-small cell lung cancer

Abstract

Non-small cell lung cancer (NSCLC) remains a leading cause of cancer death globally. More accurate and reliable diagnostic methods/biomarkers are urgently needed. Joint application of metabolomics and transcriptomics technologies possesses the high efficiency of identifying key metabolic pathways and functional genes in lung cancer patients. In this study, we performed an untargeted metabolomics analysis of 142 NSCLC patients and 159 healthy controls; 35 identified metabolites were significantly different between NSCLC patients and healthy controls, of which 6 metabolites (hypoxanthine, inosine, L-tryptophan, indoleacrylic acid, acyl-carnitine C10:1, and lysoPC(18:2)) were chosen as combinational potential biomarkers for NSCLC. The area under the curve (AUC) value, sensitivity (SE), and specificity (SP) of these six biomarkers were 0.99, 0.98, and 0.99, respectively. Potential diagnostic implications of the metabolic characteristics in NSCLC was studied. The metabolomics results were further verified by transcriptomics analysis of 1027 NSCLC patients and 108 adjacent peritumoral tissues from TCGA database. This analysis identified 2202 genes with significantly different expressions in cancer cells compared to normal controls, which in turn defined pathways implicated in the metabolism of the compounds revealed by metabolomics analysis. We built a fully connected network of metabolites and genes, which shows a good correspondence between the transcriptome analysis and the metabolites selected for diagnosis. In conclusion, this work provides evidence that the metabolic biomarkers identified may be used for NSCLC diagnosis and screening. Comprehensive analysis of metabolomics and transcriptomics data offered a validated and comprehensive understanding of metabolism in NSCLC.

Fig 1. An overview workflow of the comprehensive analysis of metabolomics and transcriptomics in NSCLC.

Fig 2

Representative TIC chromatograms of serum from NSCLC patient (red) and healthy people (blue) in ESI+ or ESI- mode.

Fig 3. OPLS-DA score plots, S-plots, and validation plots for the metabolic profiling results of NSCLC patients and healthy people.

(A) OPLS-DA score plot for NSCLC patients versus healthy controls in the ESI+ mode (R2X = 0.370, R2Y = 0.915, Q2 = 0.855). (B) OPLS-DA score plot for NSCLC patients versus healthy controls in the ESI- mode (R2X = 0.369, R2Y = 0.904, Q2 = 0.816). (C) S-plot for NSCLC patients versus healthy controls in the ESI+ mode. (D) S-plot for NSCLC patients versus healthy controls in the ESI- mode. (E) Permutation test for NSCLC patients versus healthy controls in the ESI+ mode. (F) Permutation test for NSCLC patients versus healthy controls in the ESI- mode. The criteria for stability and credibility are as follows: all permuted R2 and Q2 values on the left are lower than the original point on the right, and the Q2 regression line in blue has a negative intercept.

Fig 4

ROC curves of potential metabolic biomarkers for the discovery set (A) and validation set (B); Y-predicted scatter plot (C) of the validation set.

Fig 5

Venn plot (A) of metabolic pathways enriched by metabolomics and transcriptomics; pathway analysis plot (B) of metabolites with a significant difference between samples from NSCL patients and healthy controls; connected network (C) of metabolites and genes of metabolomics and transcriptomics analysis. The quadrangle in blue indicates differential metabolites, and the nodes in green or red indicate increased or decreased expression of genes.