Algorithm Development Of Chromatographic Peak Detection For UPLC-MS Metabolomics Data And Study On The Regulation Of Scutellaria Barbata On Lung Adenocarcinoma Cell

Metabolomics,as a comprehensive and systematic study on the relationship between endogenous small molecular metabolites and physiological/pathological changes of the body,can reveal the internal changes of the body,which has gradually become a powerful tool to reveal the pathogenesis of complex diseases and to illustrate the mechanism of TCM regarding to its multiple components and multiple targets.Nuclear magnetic resonance（NMR）has the unique advantages of simple sample preparation,non-biased detection,excellent reproducibility and structure identification and elucidation.Ultra High Performance Liquid Chromatograph-Mass Spectrometry（UPLC-MS）can detect and cover large number of metabolites because of its powerful capacity of chromatographic separation and the high resolution and sensitivity of mass spectrometry detection.NMR and LC-MS are the complementarily and most widely used metabolomics techniques.Unfortunately,so far UPLC-MS based metabolomics is confronted with the inconsistent results of peak detection and quantification,different algorithms have different coverage and accuracy because of the existence of inconsistent false positives and false negatives and different definition of peak data range.A new‘shape-orientated’continuous wavelet transform（CWT）-based algorithm employing an adapted Marr wavelet（AMW）with a shape matching index（SMI）,defined as peak height normalized wavelet coefficient for feature filtering,was developed for chromatographic peak detection and quantification.Taking the complementary advantages of NMR and UPLC-MS,using DHA as the positive control and non-small cell lung cancer（NSCLC）adenocarcinoma PC9 cell as the cell model,we jointly applied NMR and UPLC-HRMS to carry out the metabolomic effect of S.barbata on the lung adenocarcinoma PC9 cell,and preliminarily explore the metabolic regulation mechanism of S.barbata on the lung adenocarcinoma cell.Objective:To explore and understand why different peak detection algorithms generate inconsistent peak lists and significantly different peak quantitation,and to develop peak detection algorithm that can reduce both false positives and false negatives and accurately quantify chromatographic peaks.Further,NMR and UPLC-MS were jointly used to study the metabolomics of PC9 cell treated by S.barbata,to explore and understand the mechanism of the regulation of S.barbata on lung adenocarcinoma cell.Methods:（1）An adapted Marr wavelet（AMW）was used for continuous wavelet transform and an shape matching index was introduced for peak filtering to reduce false positives.Meanwhile,the algorithm AMW-SMI alleviated the requirements on the signal intensity,and thus it enables the detection of low-intensity features and can increase the detection sensitivity and coverage,while keeping false positives low by filtering out poorly shaped traces.AMW-SMI identifies and validates ridge and valley lines in the wavelet space to determine the peak position and peak boundary to improve quantitative accuracy and stability.Compared with XCMS（cent Wave）and MZmine 2（ADAP）,four different UPLC-MS data were used to evaluate the performance of the algorithm regarding to recall,accuracy and F score.（2）The metabolic differences of lung adenocarcinoma PC9 cell between the S.barbata and the blank（negative control）groups were analyzed based on ¹H NMR and UPLC-MS metabolomics.An in-house AU program of Top Spin 4.0.7,the Data Analysis 4.0 and the developed AMW-SMI algorithm were used for high-throughput processing of ¹H NMR spectra and UPLC-MS data,respectively.Based on the ¹H NMR data,the probability quotient normalization（PQN）factors were calculated using an in-house R script.Pareto scaling was used to standardize the PQN normalized¹H NMR and the UPLC-MS data.Multivariate（using SIMCA-P 14.0）and univariate statistical analysis were applied on the regularized data matrices,giving the statistically significant differential variables.Differential metabolites were identified by querying the relevant NMR（chemical shifts）and LC-MS/MS（product ion fragments）data against the public metabolomics database（HMDB and BMRB,etc.）.Metabo Analyst was used to analyze the metabolic pathways associated with the identified differential metabolites to illustrate the metabolic pathways closely related to S.barbata.Similar procedure was performed to illustrate the metabolic pathways closely related to DHA,and comparison was made with those related to S.barbata.Results:（1）A new‘shape-orientated’continuous wavelet transform（CWT）-based algorithm employing an adapted Marr wavelet（AMW）with a shape matching index（SMI）,defined as peak height normalized wavelet coefficient(′₈（6）/)for feature filtering,was developed for chromatographic peak detection and quantification,which can reduce both the false positives and false negatives and accurately quantify chromatographic peaks.The performance of AMW-SMI is evaluated qualitatively（by recall,precision and F-score）and quantitatively（by ratio of iso-topic features and triplicate RSD）on the LC-MS data of model mixtures of 21 human metabolite standards and 8 plant metabolite standards,and of serum sample spiked with the 21 human metabolite standards,and on the triplicate LC-MS data of the same sample of cell metabolomic extracts.Compared with XCMS（cent Wave）and MZmine 2（ADAP）,the proposed AMW-SMI algorithm can faithfully identify chromatographic peaks with significantly fewer false positives and demonstrated general superiority in terms of qualitative precision（robustness）and quantitative accuracy（by ratio of isotopic features）,and comparable recall（sensitivity）and quantitative stability（by RSD of triplicate measurements）.（2）Multivariate and univariate statistical analysis（MVA and UVA）of the ¹H NMR metabolomics data of the blank（negative control）and S.barbata treated groups of lung adenocarcinoma PC9 cell revealed.16 differential metabolites（Glycerophosphocholine,Betaine aldehyde,L-threonine,L-alanine,L-serine,L-asparagine,L-isoleucine,Uridine,L-leucine,L-proline,Choline,Trans-4-hydroxy-L-proline,Glutathione,ATP,Myo-inositol and L-glutamic acid）related to S.barbata.Similarly,MVA and UVA of the UPLC-HRMS metabolomic data of the blank and S.barbata treated groups revealed another set of 16 differential metabolites（Choline,Phosphorylcholine,L-leucine,Pantothenic acid,Uridine,L-Pipecolic acid,Trigonelline,Glutamamide,N-Acetyl-L-aspartic acid,L-methionine,Citrulline,Glutathione,Betaine,L-glutamine,Creatine and Argininosuccinic acid）related to S.barbata.Combining the results from NMR and LC-MS,total 28 differential metabolites related to S.barbata were obtained,out of which18（Glycerophosphocholine,Betaine aldehyde,L-threonine,L-alanine,L-serine,L-asparagine,L-isoleucine,Uridine,L-leucine,L-proline,Choline,Trans-4-hydroxy-L-proline,Phosphorylcholine,Pantothenic acid,L-Pipecolic acid,Trigonelline,Glutamamide and N-Acetyl-L-aspartic acid）were up-regulated and the other 10（Glutathione,ATP,Myo-inositol,L-glutamic acid,L-methionine,Citrulline,Betaine,L-glutamine,Creatine and Argininosuccinic acid）were down-regulated.By the same procedure,32 differential metabolites（ATP,Betaine aldehyde,Creatine,Glycerophosphocholine,Glycine,L-alanine,L-asparagine,L-glutamic acid,L-glutamine,Glutathione,L-leucine,L-isoleucine,L-serine,L-threonine,L-tyrosine,L-valine,Myo-inositol,NAD,Uridine,L-phenylalanine,L-methionine,L-carnitine,L-tryptophan,L-histidine,choline,Pantothenic acid,Citrulline,Phosphorylcholine,Betaine,Adenine,Nicotinic acid and Nicotinamide）related to DHA were illustrated.Referring to DHA,21 of the 28 different metabolites related to S.barbata were same with those related to DHA,15（Glutathione,Betaine aldehyde,L-threonine,L-serine,L-asparagine,L-isoleucine,ATP,Myo-inositol,L-leucine,L-glutamic acid,Choline,Phosphorylcholine,Betaine,L-glutamine and Creatine）of which had same changing trend,and 6（Glycerophosphocholine,L-alanine,Uridine,L-methionine,Citrulline and Pantothenic acid）of which had the opposite changing trend.（3）Based on the 28 differentially metabolites related to S.baristoria,7 metabolic pathways associated with S.baristoria were identified by Metabo Analyst.Similarly,11 metabolic pathways related to DHA were obtained.Referring to DHA,6 of the 7metabolic pathways associated with S.barbata are same with those associated with DHA,namely glutamine and glutamate metabolism,arginine biosynthesis,alanine,aspartic acid and glutamate metabolism,glycine,serine and threonine metabolism,glutathione metabolism,and aminoacyl t RNA synthesis.Conclusion:In this paper,a new shape-oriented peak detection algorithm（AMW-SMI）is developed,which can improve the accuracy and stability of chromatographic peak quantification while significantly reducing false-positive and false-positive peaks.Integrating ¹H NMR and UPLC-MS metabolomics,it revealed that S.barbata and DHA could probably be involved in the same six metabolic pathways,which implied that S.barbata and DHA may regulate the metabolism of PC9 cell and affect the cell proliferation through the same six metabolic pathways.