樣本信息:

最初，研究團隊對每組的血液樣本進行處理并檢測，得到數據并進行總體分析：

Fig 1. Scheduled multiple-reaction monitoring (S-MRM) chromatograph of ?ve metabolites in positive ion mode. Full, dash and dot line stand for health control, ESCC patients and mESCC patients respectively.

從多維統計分析圖中看出不同組樣本間有明顯的分離趨勢。
然后將轉移組和非轉移組食管癌病人血清樣本數據與正?？刂平M樣本數據今天進行對比分析，找出差異物質，并結合數據庫進行結構鑒定。

以上表格為淋巴結未轉移組和淋巴結轉移組病人分別于健康控制組對比得到的差異化合物，總體可用如下圖表示：

疾病組（淋巴結轉移和非轉移組）與健康組之間對比有很多代謝物發生了變化，其中actic acid和fatty acids,升高，glucose、glutamine和TCA循環中的代謝物含量降低，這在癌癥中是一個很常見的現象。另外支鏈氨基酸的代謝產物如：2-ketoisovaleric acid, 2-ketoisocap- 260 roic acid, 和3-methyl-2-oxovaleric acid的含量顯著降低。另外，cysteine, methylcysteine, pyrogallol, tocopherol等和氧化應激有關的代謝物的含量明顯降低。此外tryptophan, indolelactic acid, uric acid, p-cresol, phosphoethanolamine, 和 cholesterol等物質在疾病組含量明顯降低，而α-hydroxybutyric acid, aspartic acid, β-alanine, 和 maltose的含量則顯著升高，這些現象不僅可以看出代謝的變化同時也反映一些潛在的病理機制。
接著，研究團隊又對疾病組（ESCC&mESCC）組進行對比分析：

Figure 2. Discriminating plots of non-metastatic and metastatic ESCC patients: (A) scores plot of the PLS-DA model, (B) plot of the permutation test (200 times) of the PLS-DA model, (C) scores plot of the OPLS-DA model, and (D) scores plot of the prediction analysis of the OPLS-DA model.

可以看出在模型較好的前提下，疾病組之間也有很大的差異。接著對這兩組數據進行對比分析，并對篩選出的差異物進行結構鑒定：

可以看出在淋巴結轉移組中氨基酸及其衍生物如glutamine, cystine, tryptophan,γ-amino-butyric acid,valine,leucine,和pyrrole-2-carboxylic acid等含量明顯降低，而glutamic acid則在淋巴結轉移組含量升高。
以上物質大多已經發現和癌細胞的增殖、調亡、免疫逃跑、轉移以及氧化壓力等有關，可總結為如下圖：

Figure 3. Heatmap and function classi?cation of 15 di?erential metabolites between non-metastatic and metastatic ESCC patients. Heatmap (left) was produced by average normalized peak areas with z- score scaling of healthy controls (C), non-metastatic ESCC patients (E), and metastatic ESCC patients (M). These metabolites showed progressive elevation or decline with the progression of ESCC (from C to E to M). In total, 13 metabolites were assigned to ?ve function groups (right), except for iminodiacetic acid and glycolic acid. The green background indicates that the function is improved in metastatic ESCC patients because of the metabolites with changed levels (left) compared to that in non-metastatic ESCC patients, whereas the functions classi?ed with a red background represent the opposite of this.

為了在這些兩組之間的差異代謝物之間找出用于診斷淋巴結是否轉移的生物標志物，研究團隊對這些代謝物運用相關算法進行計算，通過training set和test set篩選出特異性和靈敏度最高的診斷物質。

Figure 4. Box plots of serum valine, γ-aminobutyric acid, and pyrrole-2-carboxylic acid, ROC curves based on the binary logistic regression model by the combination of three serum metabolites, and their prediction plots based on the optimal cuto? value from ROC curves. (A) Values in the box plots are shown as the normalized peak areas of the metabolites in healthy subjects (healthy, green), non-metastatic patients (ESCC, blue), and metastatic patients (mESCC, red). (B) The ESCC samples from the training set were applied to construct a binary logistic regression model based on the combination of serum valine, γ-aminobutyric acid, and pyrrole-2-carboxylic acid, and the ROC curves of the training set (B, left) and test set (B, right) were obtained from the above established prediction model. (C) The optimal cuto? value with the highest sensitivity and speci?city in the ROC curves of the training set was obtained (0.558) and applied to evaluate the prediction capacity (85 and 93.3% for test set and training set, respectively) of the current model, where 0 and 1 on the x axis represent ESCC and mESCC patients, respectively, and red diamonds represent samples.

最后篩選出三個物質作為鑒別淋巴結是否轉移的“指示器”，分別為Valine, GABA和Pyrrole-2-carboxylic acid。