Fig. 5: A description of metabolites emphasized by TMR-CT. | npj Precision Oncology

Fig. 5: A description of metabolites emphasized by TMR-CT.

From: Deep representation learning of tissue metabolome and computed tomography annotates NSCLC classification and prognosis

Fig. 5

a Correlation of performance of the TMR-CT model developed by DPCCA, the bar plot shows the highest absolute correlation between metabolites and the TMR-CT features; we expand the top ten least correlated metabolites and the top ten most correlated metabolites in the yellow boxes. b Unsupervised hierarchical clustering of the TMR-CT features with the metabolomic profile from nodule UHRS identified three distinct subgroups. The blue rectangle contains a summary of the performance of the distinct cluster for classification and prognosis. The pentose phosphate metabolite sedoheptulose-7-phosphate, an important source for ribonucleotides and reduced nicotinamide adenine dinucleotide phosphate (NADPH), adenosine, reported in our previous lung metabolomics publication to be high in tumor tissue are emphasized in both the highly correlated metabolite set (a) and cluster 2 (b). Other metabolites such as asparagine, an important regulator of cancer cell amino acid homeostasis, anabolic metabolism and proliferation are emphasized only by the clustering approach19,45,46.

Back to article page