Identification of potential metabolic biomarkers of rectal cancer and of the effect of neoadjuvant radiochemotherapy

Figures

• 

  Fig 1. Plasma metabolic alterations in patients with Rectal Cancer (RC). (A) Schematized view of differences in metabolites related to the glycolytic pathway, citric acid cycle, and amino acid pathways between control and RC patients. Colours indicate statistical comparisons as shown in the legend. (B) Average fold changes of metabolite concentrations expressed as bubble plots. The bubble sizes estimate the magnitude of change. Colours represent a significant statistical increase (red), decrease (blue) or not significant (grey). NRCT: Neoadjuvant radiochemotherapy.

• 

  Fig 2. Plasma metabolic alterations in patients with Rectal Cancer (RC) after Neoadjuvant Radiochemotherapy (NRCT). (A) Schematised view of differences in metabolites related to the glycolytic pathway, citric acid cycle, and amino acid pathways between control and RC patients. Colours indicate statistical comparisons as shown in the legend. (B) Average fold changes of metabolite concentrations expressed as bubble plots. The bubble sizes estimate the magnitude of change. Colours represent a significant statistical increase (red), decrease (blue) or not significant (grey).

• 

  Fig 3. Multivariate analysis of plasma metabolites patients with Rectal Cancer (RC) and the control group pre- and post-Neoadjuvant Radiochemotherapy (NRCT). From left to right, partial least square discriminant analysis (PLSDA), variable importance in projection (VIP) scores of the PLSDA and heatmap analyses comparing RC patients and control group (A) and RC patients pre- and post-NRCT (B).

• 

  Fig 4. Spearman pairwise correlation matrices of metabolites and Paraoxonase-1 (PON1) activity and concentration pre- and post- Neoadjuvant Radiochemotherapy (NRCT). The magnitude and direction of the correlation of metabolites concentration pre- (A) and post- (B) treatment are shown by the circle size (larger is stronger) and colours (positive correlation in dark blue and negative correlation in pink), respectively. a = p<0.05; b = p<0.01; c = p<0.001.

• 

  Fig 5. Serum Paraoxonase-1 (PON1) and α-ketoglutarate concentrations discriminate between patients with Rectal Cancer (RC) and the control group. PON1 and α-ketoglutarate concentrations were significantly lower in patients than in healthy subjects. Receiver operating characteristics curve-based model evaluation and confusion matrix showed that the combination of these parameters was able to efficiently distinguish both groups. AUC: Area under the curve. a = p<0.001.

• 

  Fig 6. Selected metabolites and Paraoxonase-1 (PON1) concentrations are associated with clinical variables before treatment. (A) Receiver operating characteristics curve-based model evaluation showing that the combination of fumarate, leucine and isoleucine was able to efficiently distinguish tumour location near to (<5cm) or far from (>10cm) the anal margin in patients with rectal cancer. Confusion matrix showing the performance summary of the predicted results. (B) Receiver operating characteristics curve-based model evaluation showed that PON1 concentration was able to discriminate between patients with Nodes 0+1 vs. Node 2. Sensitivity and specificity are shown as percentages. AUC: Area under the curve. a = p<0.05; b = p<0.01.

• 

  Fig 7. Selected metabolites predicting the prognosis of patients with rectal cancer before treatment. Receiver operating characteristics curve-based model evaluation showing that valine was related to Dworak grade (A), β-hydroxybutyrate, α-ketoglutarate were related to neoadjuvant radiochemotherapy (NRCT) toxicity grade (B), add succinate and glutamate were related to tumour relapse (C). Sensitivity and specificity are shown as percentages. AUC: Area under the curve. a = p < 0.05; b = p < 0.01.

• 

  Fig 8. Correlation-based network of metabolite concentrations from control subjects and patients with Rectal Cancer (RC) pre- and post-neoadjuvant radiochemotherapy. The Spearman correlation was applied to estimate correlation coefficients, and threshold tests were used to identify significant correlations with a 0.05 False Discovery Rate cutoff. Metabolites are shown as circle nodes with three different colors (pink, green and blue), depending on the group. Positive and negative correlations are illustrated as blue or red edges, respectively. Strongest correlations are visualized with thick lines (A). Metabolite Sets Enrichment analysis from metabolite concentrations. Dot plots were used to represent the Quantitative Enrichment Analysis. The size and color of the dots show the count of enriched metabolites in the pathway and the importance of the pathway enrichment, respectively (B).