Metabolomics

1
Metabolomics

• Bob Ward
• German Lab
• Food Science and Technology

2
Genome- .All the DNA Transcriptome- .All the
mRNA Proteome- .All the proteins Metabalome .All
the metabolites

Metabolomics is a post genomic technology which
seeks to provide a comprehensive profile to all
the metabolites present in a biological sample.
(Taylor et. al, 2002)
3

Limitations of ohmics technologies

• Genomics
• Static picture Expensive Not for
  individuals
• Transcriptomics
• Need Genome (annotations) Correlated with
  proteome? Sampling issues splicing
  No info on modifications
• Proteomics
• Technologically challenging Need genome?
  

4
Metabolome

• Same metabolites for all organisms
• 1k for organism vs 10k(genes) or 100k(proteins)
• Technology exists and is not too expensive
• Carbohydrate and Lipid info

5

6
Goal Discrimination between related genotypes of
Arabidopsis

• between Co10 and C24 (parent strains)
• between Co10 x C24 and progeny (F1)
• between (Co10 x C24) and (C24 x Co10)
• -Maternal line donates both mitochondria and
  chloroplast
• -Clear-cut realization of effectiveness
• -Potential to uncover biologically relevant info

7
Instrumental and Informatic Tools

• GC/MS-Separation/identification of polar
  metabolites in 1200 second run time
• AMDIS deconvoluting software
• MassLab to choose target ions
• R for statistics
• WEKA (standard neural network approach)
• Euclidean distance
• Principal Component Analysis

8
Data Work-Up

• Selection of reference chromatogram (F1)
• 8 individual samples for each genotype
• no replicates
• Selection of target peaks/analytes (433)
• normalized (mg analyte/wt sample)to internal
  standard (ribitol)
• Allows for simple 2-D matrix

9

• 201 metabolites identified in some detail
• (92 as molecular type and 109 by chemical
  property)
• High variance in low numbers corresponds to core
  metabolites

10
Co10 1-8 C24 9-16 Co10 x C24 17-24 C24 x
Co10 25-32
11
Neural Network Analysis
P0.27
Lack of samples precluded use of a training
subset Leave one out cross used for
training Model judged by ability to classify
remaining object (repeated for all
objects) Allows for maximal use of data for
validation when n is low
12
Clustering by Euclidean distance
Co10 1-8 C24 9-16 Co10 x C24 17-24 C24 x
Co10 25-32
13
Principal Component Analysis

• Used to tease out role of individual metabolites
  in discrimination
• Unsupervised multivariate analysis applied to
  functions of many attributes
• Transformation of large set of related values to
  smaller set of uncorrelated variables
• Attempts to express maximum variance in data
• PCs are axes in multidimensional space
• Object characterized by distance to axis

14
Variance of data explained by first few principal
components
PCA algorithm from MatLab 78 of variation of
data from first 3 PCs
15
Principal Component Analysis
Co10 and C24 differentiated except outlier F1
genotypes cluster together
16
Contribution of each variable to first PC
Malate and Citrate- metabolites of TCA cycle
17
Relative peak area for metabolites malate and
citrate
Co10 contains outlier..may explain
misclassification
18
Other significant results

• Parental genotype removed from PCA analysis and
  F1s discriminated by glucose and fructose
• Inference that the first PC differentiates
  parental line, and 2nd and 3rd differentiate F1
• Malate and Citrate from TCA, glucose and fructose
  from chloroplasts

19
Conclusions

• Advances in technology will improve detection
  limits and will allow characterization of
  metabolites
• Formalized ontology needed to link chemical
  structure with pathways
• Metabolite profiling is an exciting new field
  which complements other non-hypothesis driven
  global analysis technologies
• Large amounts of informatic support to develop
  field and to correlate data from genomics,
  microarrays, and proteomics