Multiple knockout analysis of genetic robustness in the yeast metabolic network

1
Multiple knockout analysis of genetic robustness
in the yeast metabolic network
David Deutscher, Isaac Meilijson, Martin Kupiec
Eytan RuppinNature genetics, Vol 38, No 9,
Sep 2006

• Byoungkoo Lee
• Computational Biology
• Carnegie Mellon University

2
Outline

• Background
• General Ideas
• What the other scientists have figured out so
  far?
• What is a challenging problem for the authors?
• How did they tackle the problem?

3
Outline

• Background
• Method
• Genetic Robustness k Robustness
• Metabolic Pathway
• Computing Procedure
• Flux Balance Analysis

4
Outline

• Background
• Method
• Conclusions Results

5
Outline

• Background
• Method
• Conclusions Results
• Discussions
• Contributions
• Critiques
• Future Problems

6
Outline

• Background
• Method
• Conclusions Results
• Discussions
• References
• Acknowledgements

7
Background

• Some genes are essential for a cell to grow,
  while some other genes are not.

8
Background

• Some genes are essential for a cell to grow,
  while some other genes are not.
• These essentialities are good methods to find the
  function of each gene.
• Single gene knockout test

9
Background

• Single Gene Knockout test (experiment)
• Giaever et al,
• Nature 2002

10
Background

• Some genes are essential for a cell to grow,
  while some other genes are not.
• These essentialities are good methods to find the
  function of each gene.
• To find a interaction between genes, multiple
  knockouts tests are needed.
• Genetic Robustness
• Functional backup interactions between genes

11

• Gene Function network
• (1,755 attribute pairs among 285,390 different
  Gene Ontology attributes, Tong et al. Science
  2004)

12
Background

• Some genes are essential for a cell to grow,
  while some other genes are not.
• These essentialities are good methods to find the
  function of each gene.
• To find a interaction between genes, multiple
  knockouts tests are needed.
• However, the multiple gene knockouts experiments
  for the large-scale network are not easy.
• Double knockout tests were done experimentally to
  study small-scale networks.
• (Segre et al. Nature genetics, 2005)

13
Background

• Some genes are essential for a cell to grow,
  while some other genes are not.
• These essentialities are good methods to find the
  function of each gene.
• To find a interaction between genes, multiple
  knockouts tests are needed.
• However, the multiple gene knockouts experiments
  for the large-scale network are not easy.
• Multiple knockouts test in silico is one possible
  way to find the interaction between genes.
  Computational Method!

14
Genetic Robustness

• How well an organism can survive in difficult
  circumstances such as different environments and
  mutation or deletion of a gene.
• Duplication or Overlap
• If they found isoenzymes
• Alternative biochemical pathway
• If they did not found isoenzymes,

15
k Robustness

• k robustness
• (The depth of backup interaction)
• the size k of the smallest essential gene set
  that includes the knocked-out gene
• 1-robust the knockout of an essential gene
• 2-robust the knockout of any nonessential gene
  which is involved in a synthetic lethal pair

16
Metabolic Pathway

• Green Mitochondria
• Orange Cytosol
• Blue Extracelluar
• PRO L-Proline
• ARG L-Arginine
• ORN L-Ornithine
• GLU L-Glutamate
• NGLUSN-Acetyl-L-glutamate 5-semialdehyde
• NORNN2-Acetyl-L-ornithine
• NGLUPN-Acetyl-L-glutamate 5-phosphate
• NGLUN-Acetyl-L-glutamate

• Proline and Arginine metabolism,
• Deutscher et al. Nature genetics, 2006
  (supplementary Fig 1.)

17
Metabolic Pathway

• Green Mitochondria
• Orange Cytosol
• Blue Extracelluar
• PRO L-Proline
• ARG L-Arginine
• ORN L-Ornithine
• GLU L-Glutamate
• NGLUSN-Acetyl-L-glutamate 5-semialdehyde
• NORNN2-Acetyl-L-ornithine
• NGLUPN-Acetyl-L-glutamate 5-phosphate
• NGLUN-Acetyl-L-glutamate

• Proline and Arginine metabolism,
• Deutscher et al. Nature genetics, 2006
  (supplementary Fig 1.)

18
Metabolic Pathway

• Green Mitochondria
• Orange Cytosol
• Blue Extracelluar
• PRO L-Proline
• ARG L-Arginine
• ORN L-Ornithine
• GLU L-Glutamate
• NGLUSN-Acetyl-L-glutamate 5-semialdehyde
• NORNN2-Acetyl-L-ornithine
• NGLUPN-Acetyl-L-glutamate 5-phosphate
• NGLUN-Acetyl-L-glutamate

• Proline and Arginine metabolism,
• Deutscher et al. Nature genetics, 2006
  (supplementary Fig 1.)

19
Computing Procedure
20
Computational Method

• Yeast Database
• Focusing on the 484 genes
• Known ORFs
• Not on a dead-end pathway
• Input parameters Environments
• Minimal media glucose, oxygen, ammonia,
  phosphate, sulfate and potassium
• Rich media Minimal media, 20 amino acids,
  purines and pyrimidines

21
Computational Method

• Input parameters of Knockout genes
• Testing all combinations of up to four knockouts
• Testing all combinations of five knockouts are
  not easy. 2 years! ? (a cluster of ten computers)
• For example, the number of all combinations of
  five knockouts among 484 genes 2.171011
• For example, the number of all combinations of
  four knockouts among 484 genes 2.26109
• Using stochastic sampling methods for more than
  four knockouts. 2 weeks! ?

22
Computational Method

• FBA (Flux Balance Analysis)
• Useful technique for analysis of metabolic
  capabilities of cellular systems.
• Based on mass balances around intracellular
  metabolites.
• Find an upper bound on the growth rate of the
  organism.
• Linear optimization is used to calculate optimal
  growth rates for objective functions such as
  maximization of biomass production or
  minimization of nutrient utilization.

23
(No Transcript)
24
FBA (Flux Balance Analysis)

• Stoichiometric Matrix S
• Flux Matrix V
• SV 0 in Steady State

25
Linear Programming

• Objective Function
• Max Biomass Production
• Max Cell Growth
• Constraints
• Flux Balance Constraints
• SV 0
• Capacity Constraints
• 0 Vi
• a Vj b

26
Linear Programming

• Gene knockout
• Solution Space will be reduced.
• Different Environments
• Solution Space will be changed.

27
1st Result
28
1st Conclusion

• 19 of the genes in yeast are essential in
  laboratory condition in vivo. (Giaever, et al.
  Nature, 2002)
• Found 48 essential genes, 14 essential pairs, 17
  triplets, and 39 quadruples (by exhaustive
  multiple knockout search in silico)
• Additional 173 contributing genes more than 4
  knockout case

29
Distribution of Back up mechanisms By
alternative pathway black (a,e)By duplication
light gray (c,g)By both dark gray (b,f)
2nd Result (Gene Histograms)
30
2nd Conclusion

• Minimal medium, Single gene deletion
• Genetic Robustness from the duplication of a
  specific gene
• Rich medium, multiple gene knockouts
• Higher depths and more complex
• Genetic Robustness from alternative pathways

31
Discussions

• Contribution
• Multiple knockouts test more than 2 for
  large-scale network
• Random sampling test more than 4
• Critique
• Optimistic bias (upper bound from FBA)
• falsely predicting viability gt falsely predicting
  lethality
• For big k-rubustness, the results can be wrong.
• more than 9

32
Discussions

• Future Problems
• Considering multiple experimental results
• DNA microarray, Protein microarray
• Considering different analysis tool and more
  efficient algorithms
• avoid an optimistic bias
• test multiple knockout more than 9

33
References

• Deutscher et al. Multiple knockout analysis of
  genetic robustness in the yeast metabolic
  network, Nature genetics, Vol 38, Sep 2006
• Giaever et al. Functional profiling of the
  Saccharomyces cerevisiae genome, Nature, Vol
  418, July 2002
• Tong et al. Global Mapping of the Yeast Genetic
  Interaction Network, Science, Vol 303, Feb 2004

34
Acknowledgements

• Dr. Schwartz
• Dr. Cohen
• All My Friends in Computational Biology Program
  and in Dr. Schwartzs Lab