Correlation of analytical data obtained by
NMR and LC-MS/MS in metabolomics studies
Chrysovalantou Anastasia Chatziioannou1, Dimitris Palachanis2, Christina Virgiliou1,
Panagiotis Zoumpoulakis3, Eleni Gika2, Apostolos Athanasiadis4, Georgios Theodoridis1
contact info:
Chatziioannou
Chrysovalantou
chatzana at
chem.auth.gr
gr.linkedin.com/
in/cchatziioannou
Department of Chemistry, Aristotle University of Thessaloniki, Greece / 2 Department of Chemical Engineering, Aristotle
University Thessaloniki, Thessaloniki, Greece / 3 Institute of Biology, Medicinal Chemistry & Biotechnology, National Hellenic
Research Foundation, Greece / 4 Obstetrics Clinic, Papageorgiou Hospital, Aristotle University Thessaloniki, Greece
1
In metabolomics a multitude of methods is typically used:
NMR, LC­MS and GC­MS but data is only partially
correlated usually by manual comparison of findings (e.g.
comparing signals for a few selected metabolites). At
present there is no technology in linking such complex
multi-dimensional data, to combine them into one table of
metabolites. Generation of such tools will advance the
process and the success rate of discovery work.
Blood serum and human amniotic fluid (HAF) samples
were collected from 29 women that underwent
amniocentesis for prenatal diagnosis. Samples were
analysed by two different and complementary analytical
techniques (NMR and LC-MS/MS). The initial scope was the
unification and the combination of all the information
extracted from the two analytical techniques and, thus, the
better understanding of the studied clinical question.
The holistic NMR analysis was accomplished utilizing a
600 MHz Varian NMR spectrometer by applying CPMG
pulse sequence and using sodium maleate as internal
standard. Sample pretreatment of HAF samples included
freeze­drying to increase signal intensities.
The LC­MS/MS analysis was achieved by the ACQUITY
Ultra Performance Liquid Chromatography System Xevo
TQD MS System (Waters). An ACQUITY HILIC,
2.1x150 mm, 1.7 um, BEH amide column together with a
ACQUITY UPLC BEH Amide 1.7 um VanGuard pre­
column was used under 500 uL/min. Gradient elution
employed a ramp of water vs ACN both buffered with
formic acid and ammonia (10 mM). 100 MRM channels
were set in time windows of ca.1­3 min for total analysis
time of 40 min (Virgiliou et al, Electrophoresis, Submitted).
Identified metabolites of each technique were integrated
using Chenomx (identification) and MestReNova 9.1
(integration) software for the 1H-NMR data and Waters
MassLynx XS software for the LC-MS/MS data. The NMR
signals were normalized to the signal of sodium maleate
per sample.
An effort to investigate the existence of correlation
between the integrated areas of the metabolites
identified in both techniques was made. In that way, the
correlation of the two techniques would be proven. Only
after that procedure the extracted information of the
areas of unique metabolites from each technique could
be combined, in order to form a single data matrix (peak
table) containing the entire extracted information.
2"Hydroxybutyrate/
2"Hydroxyvalerate/
3"Hydroxyisobutyrate/
3"Hydroxybutyrate/
5,6"Dihydrothymine/
Acetate/
Acetoacetate/
N"acetylornithine/
Acetone/
Succinate/
Citrate/
2"Oxoglutarate/
Dimethyl/sulfone/
Methanol/
Mannose/
Maleate/
Fumarate/
π"Methylhis5dine/
Formate/
Valine/
/
2"Hydroxyisovaleric/
acid/
Alanine/
Crea5ne/
Crea5nine/
Glucose/
Glutamine/
Glycine/
His5dine/
Isoleucine/
Lac5c/acid/
Leucine/
Lysine/
Methionine/
Phenylalanine/
Pyruvic/acid/
Threonine/
Tyrosine/
2"Hydroxyisobutyric/acid/
Acetylcarni5ne/
Adenosine/
Allose/Mannose/Galactose/
Arginine/
Asparagine/
Betaine/
Caffeine/
Choline/
Citrulline/
Co5nine/
Cys5ne/
Dimethylamine/
Dulcitol/
Fructose/
Glutamic/acid/
Hippuric/acid/
Malic/acid/
Mannitol/
Nico5namide/
Norvaline/Valine/
Ornithine/
Pantothenate/
Proline/
Pyroglutamic/acid/
Ribose/
Serine/
Spermidine/
Taurine/
Theobromine/
Thiamine/
Trimethylamine"n"oxide/
Tryptophan/
Uridine/
Venn diagrams for amniotic fluid compounds
identified with each technique. This diagram
was applied to specify the metabolites identified
with both analytical techniques (common
metabolites). 17 metabolites out of the 51 and 37
metabolites identified in HAF samples with LCMS/MS and NMR respectively were common. In
serum samples the common metabolites were 17
out of the 55 and 27 metabolites identified with
LC-MS/MS and NMR respectively.
Heatmap plots for each biological sample based on the Pearson
correlation coefficient. The resulted heatmap for HAF samples
revealed low correlation between the common integrated signals of
targeted LC-MS/MS and NMR. The resulted heatmap for serum samples
revealed higher correlation between the common integrated signals.
Nevertheless, in both matrices, there are metabolites such as leucine,
alanine, creatine and glutamine that appear to be highly correlated. On
the contrary, creatinine appears to be one of the most non-correlated
metabolites.
R2=0.700
R2=0.685
R2=0.070
Correlation of the areas of the metabolites
Val, Ala, Crtn identified with LC-MS/MS and
NMR in blood serum samples.
DISCUSSION
Combination of LC-MS/MS and NMR data is not possible for the HAF samples, as the correlation of the common metabolites
is low. This could result from the sample preparation procedure. On the contrary, high correlation is observed on the
common metabolites found in serum. This observation leads us to the next step, the fusion of the two datasets reached with
the different analytical techniques. A major obstacle in this direction is the large difference in signal counts (3 orders of
magnitude) and the nature of signal response (analyte specific in MS).
This research has been co-financed by the European Union (European Social Fund – ESF) and Greek national funds through the Operational Program "Education and Lifelong
Learning" of the National Strategic Reference Framework (NSRF) - Research Funding Program: Excellence II - Metabostandards. Investing in knowledge society through the
European Social Fund. More information can be found at http://users.auth.gr/gkikae/aristeia