| Objective: To explore the serous potential metabolic markers of a GVHD after HSCT based on GC/MS.Methods: GC/MS was used to deal with the 174 clinical samples from 61 patients which were collected right within 7 days, 14 days, 28 days, and 29 to 60 days after HSCT. AMDIS, NIST, and retention index were applied to identify the metabolites in the serum. After that, several statistics processing means were chosen to screen out the combination of metabolic markers of different time points, including OPLS-DA, t test, GA, Logistic regression, and ROC curve.Results:(1) A stable and reliable metabolomic analysis procedure was build based on GC/MS, which had already met the need of relative quantification of serum metabolites, with more than 95% area RSD below 15% of common peaks in the Total ion chromatogram in the pre-examination of the experiment.(2) 54 metabolites were identified with the help of AMDIS, NIST and retention index after analyzing TIC for all the samples.(3) Using OPLS-DA to analyze all the metabolites at each time point, the samples were obviously separated into two groups. Characteristic metabolites were selected by S-Plot and VIP as the following, urea for the time point within 7 days after HSCT,urea and hexadecanoic acid for 14 days after HSCT, urea for 28 days, while lactic acid, fructose, talose, and glycolic acid for the time point 29 to 60 days after HSCT.(4) With the utilization of independent-sample t test, several characteristic metabolites were screened out under the standard p≤0.05. They were alanine for the first time point, anhydrosorbitol and glyceric acid for the second one, glycine and anhydrosorbitol for the third time point, and lactic acid, propanoic acid, valine, urea, glycine, serine, anhydrosorbitol, glucose, myo-inositol, propylene glycol, α- tocopherol, glycolic acid, 2,3-dihydroxybutanoic acid for the last time point.(5) The same GA calculation was given to all the metabolites for 5 times, top 10 metabolites were chosen as characteristic metabolites for each time point, Alanine, 3-hydroxybutanoic acid, glycerol, phenylalanine, anhydrosorbitol, tyrosine, oleic acid, ornithine, phospate, glyceric acid for the time point within 7 days after HSCT, alanine, urea, anhydrosorbitol, uric acid, 1-monopalmitin, aspartic acid, glyceric acid, mannose, pentadecanoic acid, eicosanoic acid for 14 days after HSCT, glycerol, oleic acid proline, glycine, serine, glyceric acid anhydrosorbitol, linoleic acid, tryptophan, phospate, for 28 days after HSCT, and lactate, oxalate, valine, glycine, threonine, anhydrosorbitol, myo-inositol, propylene glycol, α- tocopherol, glycolic acid for the time point 29 to 60 days after HSCT.(6) We used logistic regression to analyze the metabolites for each time point, with the criterion of αin=0.05 and αout=0.10. Alanine, urea, tyrosine were chosen for the first time point. Anhydrosorbitol, glyceric acid, pentadecanoic acid were for the second one. Glycine, anhydrosorbitol, glyceric acid were for the third one, and lactate, oxalate, threonine, propylene glycol were for the last time point.(7) All the characteristic metabolites were gathered together for each time point, and were put into the logistic regression formula in different panels, after which a ROC curve was established. If the ROC curve met the 3 demands below, 1) p value was below 0.01, 2) AUC was above 0.85, 3) the sensitivity and specificity were both above 70%, the panel would be the final one for right this time point. The panel for each time point is alanine+ urea+ tyrosine for the time point within 7 days after HSCT, alanine+ urea+ uric acid+ anhydrosorbitol+1-monopalmitin+ aspartic acid+ glyceric acid+ pentadecanoic acid+ eicosanoic acid for 14 days after HSCT, glycine+ serine+ anhydrosorbitol+ oleic acid+ glyceric acid for 28 days after HSCT, and lactate+ oxalate+ threonine+ propylene glycol for the time point 29 to 60 days after HSCT. Among all the panels, the panel for the first time point has the best potential in predicting the occurrence of a GVHD with the highest AUC of 0.935, and 90.5% for sensitivity, 81.8% for specificity.Conclusion:(1) The metabolic model based on GC/MS is competent to predicting the occurrence of a GVHD after HSCT.(2) A panel of metabolic markers can be screened out for each time point during HSCT, which is good in forecasting and distinguishing. |
PDF Full Text Request