Human brain tumour tissues from glioblastoma multiforme, astrocytoma, meningioma, oligodendroglioma and metastatic tumours were analyzed by metabolite-metabolite correlation analysis of HRMAS ¹H NMR spectra from the eTumour database. The following metabolites were quantified using a modified LC-Model basis set: alanine (Ala), choline (Cho), creatine (Cr), lactate (Lac), glutamine (Gln), glutamate (Glu), glycine (Glyn), N-acetylaspartate (NAA), phosphocholine (PCh), phosphocreatine (PCr), taurine (tau), myo-inositol (Ino) and various lipids/macromolecules. The estimated metabolite concentrations from LCModel fittings were used in the investigation of pairwise metabolite-metabolite correlations analysis (MMCA). Pairwise metabolite-metabolite correlations can serve as an overview of metabolism and can be helpful in understanding the cellular metabolism.All the HRMAS ¹H NMR spectral data (water pre-saturated and 30ms T₂ filtered (CPMG) spectra) were downloaded from the e-Tumour database¹.The eTumour project (2004 – 2009), funded by the EU (FP6-2002-LIFESCIHEALTH 503094), involved 11 partners across Europe and Argentina. Spectra of tumour tissue samples from verified cases of glioblastoma multiforme (GBM, n=154), astrocytoma (ASTR, n=107), meningioma (MENI, n=75) , oligodendroglioma (ODG, n=37 and brain metastasis (MET, n=34) were analysed in this study. T₂ filtered (30ms in CPMG) HRMAS ¹H NMR spectra was used for PCA analysis. Spectra data was binned with 0.01ppm interval between 0.5 ppm to 4.5 ppm (AMIX). SIMCA 14 (Umetrics) was used for the PCA analysis. LCModel was used with a modified basis set to estimate the metabolite concentrations from water suppressed-spectra; alanine, choline, creatine, lactate, glutamine, glutamate, NAA, phosphocholine, taurine, myo-inositol and lipids/macromolecules were quantifiable. The pairwise metabolite-metabolite correlations were estimated by using a mixed model method we recently developed² ; it also highlights correlations with high significance at a cut-off value of P ≤ 0.001.Many positive correlations were observed between the metabolites belonging to the same biochemical pathway, and also between other metabolites in different biochemical modules.There were many negative correlations between lipids and the small-molecule metabolites involved in glycolysis, energy metabolism, membrane metabolism and glutamine and glutamate metabolism (Fig 3). A positive correlation between lactate and alanine was observed in all brain tumour types except ODG. An upregulation of glycolysis (the Warburg effect) in these tumours might be the cause for this observation. The choline plus creatine level was previously found to be negatively correlated with lipid levels in human brain tumours from in vivo MRS spectral data³. This was attributed to dilution of metabolite levels due to a heterogeneous mix of tissues, tumour grades, inflammation, hypoxia and necrosis³. Lipid and glutamine levels from in vivo ¹H MRS data had shown a negative correlation in paediatric brain tumours⁴. It has also been shown that a positive correlation between signals at 1.3ppm and lipid pseudo-droplets⁵ (and also lipids⁶) in human brain tumour tissues consisting of no-necrosis, low necrosis and high necrosis. The negative correlation between various metabolites and lipids observed in this ex vivo study shows further evidence of these effects. ODG tumours showed a set of different correlations within the metabolites when compared to the correlation data of other brain tumour types (table 1).The PCh to GPC ratio is regarded a marker of malignancy⁷. The plot of these metabolites showed a strong positive correlation in MENI, whereas in GBM and ASTR there was no correlation (table 1).HRMAS ¹H NMR spectral data of brain tumour tissues shows numerous negative correlations between the signals of metabolites and of lipids, perhaps because the tissue samples include volumes with different tumour grades, inflammation, hypoxia and necrosis. The positive correlation observed between lactate and alanine levels in all brain tumours in this study (with an exception of ODG) might be due to enhanced glycolysis. Detection of the percentage of necrosis in human brain tumours is important for estimating grade of malignancy and also the response to therapy. In future work we will therefore analyse the in vivo ¹H MRS spectra in the eTumour database and compare them with the HRMAS ¹H NMR metabolite data. MMCA can be a helpful tool in understanding tumour metabolism.