代谢组学介绍

tissues and
b, Venn diagram of 626 metabolites in tissues measured across 16 benign adjacent
prostate tissues, 12 clinically localized prostate cancers (PCA) and 14 metastatic prostate cancers (Mets)
VIP (variable importance in the projection ) values
The influence of every term in the matrix X on all the Y's. VIP is normalized so that Sum (VIP)2 = K (number of terms in the matrix X). Terms with VIP > 1 have an above average influence on Y.
Some knowledge about prostate cancer
1. Prostate cancer — the most frequently diagnosed cancer in men 2. current diagnostic methods: using a combination of digital rectal examination and measuring the levels of the enzyme PSA in the blood serum 3. limitation of current diagnosis: the features of this kind of cancer are notoriously variable among patients.
代谢组学“是通过考察生物体系（细胞、组织或生物体）受到刺激或扰动
后（如将某个特定的基因变异或者环境变化后），其代谢产物的变化或其随 时间的变化，来研究生物体系的一门科学” 许国旺 2008
Analytical plat-forms：
(1) Nuclear magnetic resonance (NMR); (2) Gas Chromatography–Mass Spectrometry ( GC-MS ); (3) Liquid Chromatography-Mass Spectrometry ( LC-MS ); etc.
Other statistic approaches, such as t test and ANOVA, are alternatives at this step.
Potential biomarker identification: standard Student’s t test or ANOVA
PCA scores plot of onset ALL and AML patients
Data analysis
(6) 有监督的模式识别方法：
利用一组已知分类的样本作为训练集，让计算机对其进行学习，获取分类的基本模型，进而 可以利用这种模型对另一组分类未知的样本进行类别识别。
常用的有监督学习方法如偏最小二乘判别分析(Partial least squares-discriminant analysis ，PLS-DA)，正交偏最小二乘判别分析，费舍尔线性判别分析…
许国旺等著. 代谢组学-方法与应用, 科学出版社，2008年第一版:第12章,146-156
主成分分析的基本思想： 对变量X进行线性变换，形成新的综合变量PC；根据实际需要选择2-3个PC进行分析，以 达到降维和简化问题的作用（多元 二元/三元） PC1=a11X1+a21X2+ … +ap1Xp PC2=a12X1+a22X2+ … +ap2Xp
判别分析思想
应变量为定性变量，且分组类型在两组以上；自变量为可测量的度量变量。计算（线性） 判别式；将自变量代入判别式，计算每个观察样本的判别Z得分，然后根据得分值对其进行归类。
PLS-DA scores plot of onset ALL and AML patients
t(2)
t(1)
The scores t, one vector for each model dimension, are new variables computed as linear combinations of the X's. They provide a summary of X that both approximate X and predict Y.
biological sample, such as a single organism …” Oliver et al., 1998
代谢组“是指基因组的所有下游产物也即最终产物的组合，这些产物是一
些参与生物新陈代谢、维持生物体正常功能和生长发育的小分子化合物，主 要是相对分子量小于1000Da的内源性小分子”
Hierarchical cluster analysis of prostate tissue samples
Screekumar A., et al. Nature, 2009 Feruary 12, 457(7231):910-914
Principal components analysis of prostate tissue samples
许国旺等著. 代谢组学-方法与应用, 科学出版社，2008年第一版:12,146-156
偏最小二乘法分析思想
对变量进行分类：设定p个因变量Y1, … , Yp和m个自变量X1, … ，Xm，对两类变量进行建模。 提取自变量的第一成分T1和因变量的第一成分U1，使T1和U1相关程度达最大, 然后建立U1和T1 的回归方程；如果回归方程未达到满意的精度，则用同样的方法提取T2和U2。 T1=w11X1+ …+w1mXm T2=w21X1+ …+w 2mXm
Breast cancer: tCho glycerophosphocholine glucose
Bathen TF, et al. Breast Cancer Res Treat, 2007;104:181–189.
2. Response assessment to chemical drugs/therapy treatments
Metabolomic profiling of prostate cancer
a, Venn diagram of the total metabolites detected across 42 prostate-related
110 matched plasma and urine samples.
Introduction of metabonomics/metabolomics
2009-06-26
The flow of the ―omics‖ sciences: genomics, proteomics, and metabolomics
Spratlin J.L. , et al. Clin Cancer Res, 2009 January 15, 15(2):431-440
Data analysis and interpretation
(5) 非监督的模式识别方法： 利用获取的样本信息，对样本进行归类，并采用相应的可视化技术直观的表达出来，不需要 有关样品分类的任何背景信息。该方法将得到的分类信息和这些样本的原始信息（如疾病的种） 进行比较，建立代谢产物与这些原始信息的联系，筛选与原始信息相关的标志物，进而考察其中 的代谢途径。 常用的非监督学习方法如主成分分析(principal components analysis), 系统聚类分析…
blue circles-benign adjacent prostate yellow squareslocalized prostate cancer red trianglesmetastatic prostate cancer
Screekumar A., et al. Nature, 2009 Feruary 12, 457(7231):910-914
What’s in a name?
Metabolome ―… refers to the complete set of small-molecule
metabolites (such as metabolic intermediates, hormones and other signalling molecules, and secondary metabolites) to be found within a
Both as a predictive measure of efficacy and a pharmacodynamic marker
Tiziani S, Lodi A, Khanim FL, Viant MR, Bunce CM, et al. PLoS ONE, 2009, 4(1):e4251
blue-benign; yellow-localized
Metabolomics “...the complete set of metabolites/low-molecularweight intermediates, which are context dependent, varying according to the physiology, developmental or pathological state of the cell, tissue, organ or organism…” Oliver 2002
Screekumar A., et al. Nature, 2009 Feruary 12, 457(7231):910-914
Metabolomic profiling of prostate cancer
Screekumar A., et al. Nature, 2009 Feruary 12, 457(7231):910-914
GC-MS
LC-MS
Metadata obtain
Total ion chromatogram
Tao X.M., et al. Anal Bioanal Chem., 2008, 391:2881-2889
Data obtain
(1) Filtering and peak detection 滤噪、峰检测 (2) Deconvolution 重叠峰解析 （3）Peak alignment 峰对齐、匹配 （4）Normalization 归一化
Spratlin J.L. , et al. Clin Cancer Res, 2009 January 15, 15(2):431-440
Clinical applications of metabolomics in oncology
1. Search early diagnostic biomarkers
Deviation of each variables from ALL (standard deviations from average)
wk.baidu.com
Blind prediction test of PLSDA model
Y-Predicted
Three major steps of metabolomics analysis
许国旺著. 代谢组学-方法与应用, 科学出版社，2008年第一版:第一章,P1-10
What’s in a name?
Metabonomics ―…measurement of the dynamic multiparametric
metabolic response of living systems to pathophysiological stimuli or genetic modification…” Nicholson et al., 1999