1 / 41

DOMINANT-EDGE PATHWAY : A Weighted Graph Algorithm for Identifying Dominant Metabolic Pathways

DOMINANT-EDGE PATHWAY : A Weighted Graph Algorithm for Identifying Dominant Metabolic Pathways. Ehsan Ullah & Soha Hassoun Department of Computer Science Kyongbum Lee Department of Chemical and Biological Engineering Tufts University. Ethanol Synthesis in E. Coli. (Wlaschin, et al., 2006).

glynis
Télécharger la présentation

DOMINANT-EDGE PATHWAY : A Weighted Graph Algorithm for Identifying Dominant Metabolic Pathways

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. DOMINANT-EDGE PATHWAY: A Weighted Graph Algorithm for IdentifyingDominant Metabolic Pathways Ehsan Ullah & Soha Hassoun Department of Computer Science Kyongbum Lee Department of Chemical and Biological Engineering Tufts University

  2. Ethanol Synthesis in E. Coli (Wlaschin, et al., 2006)

  3. R1 : A → B R2 : B → C R3 : C → D R4 : B → E + F R5 : B → 2G R6 : G → H R7 : F → H R8 : H → I A R1 R2 R3 B C D R5 R4 G F E R7 R6 H R8 I Graph Representation of Metabolic Networks

  4. Elementary Flux Mode • An elementary flux mode is a minimal set of reactions that can operate at steady state (Schuster & Hilgetag, 1994) • Metabolic flux distributions in living cells can be represented as non-negative linear combinations of elementary modes

  5. A A A R1 R1 R1 R2 R3 B B B C D R5 R4 G F E R7 R6 H H R8 R8 I I Elementary Flux Mode Analysis: An Example A → D A → E + I A → 2I

  6. A A A R1 R1 R1 R2 R3 B B B C D R5 R4 G F E R7 R6 H H R8 R8 I I Elementary Modes Can be Combined to Produce Several Possible Metabolic Flux Distributions

  7. Our Approach for Pathway Analysis • Weighted graph search for a path containing a “dominant edge” that provides the most restrictive bottleneck • Dominance defined using edge weights that represent Gibbs free energy change

  8. Electric Circuits Chemical Reactions Energy I Chemical Potential Low Chemical Potential High Chemical Potential A ∆G < 0 ∆G > 0 Spontaneous B Low Chemical Potential High Chemical Potential

  9. Why Gibbs Free Energy? • ∆G gives estimate of reaction likelihood • For most metabolic reactions ∆G ≈ 0 • Flux through an entire pathway will be heavily influenced by a smaller subset of reactions that have very negative or positive ∆G

  10. Gibbs Free Energy • Gibbs Free Energy (G) • Maximum amount of work that can be extracted from a closed system • Calculated using Group Contribution Theory (Mavrovouniotis, 1990) • Gibbs Free Energy Change (∆G) • Difference of Gibbs Free Energy of products and substrates

  11. Dominant-Edge Pathway • Dominant reactions • for a series of reactions, it has the most positive ∆G • for parallel reactions, it has the most negative ∆G • Stoichiometrically balanced pathway

  12. Weights WR4 = -0.5 WR7 = -2.5 R4 dominates R7 A R1 B R4 F R7 H R8 I Series Reactions

  13. Parallel Reactions Weights WR4 = -0.5 WR5 = -2.5 R5 dominates R4 A A R1 R1 B B R5 R4 G F R7 R6 H H R8 R8 I I 13

  14. Problem Statement Given a metabolic network graph Gm = (V, E), and starting and ending vertices s and t, find a dominant-edge pathway from s to t.

  15. Algorithm • Identify dominant-edge path from s to t • Use modified Dijsktra’s algorithm to identify dominant-edge pathways • Augment the found path to realize a stoichiometrically balanced pathway • Use modified Dijsktra’s algorithm starting with the intermediate byproducts

  16. S S A R1 R1 R1 R2 B B B R5 R4 G F E R7 R6 Intermediate Byproducts source S and target T R3 C D H H R9 R8 T T 16

  17. A S S R1 R1 R1 R2 B B B R5 R4 G F E R7 R6 Intermediate Byproducts Dominant-Edge Path from S to T with E as abyproduct. The path is not stoichiometrically balanced R3 C D H H R9 R8 T T 17

  18. Modified Dijkstra’s Algorithm • Dijkstra’s Shortest Path Algorithm • Finds shortest paths from a given source vertex vV to all other vertices in weighted graph G=(V, E), such that all edge weights are nonnegative • Modified Dijkstra’s Algorithm (Bottleneck Problem) (D.R.Fulkerson, 1966) • Selects reaction with most negative ∆G in parallel reactions • Select reaction with most positive ∆G in series reaction

  19. Example Problem Find Dominant-Edge Pathway from S to T A -24 B S -9 -18 -6 -2 E C -19 -30 -11 D -15 -5 -6 -25 -20 T F

  20. Example Problem S = { } Q = { S, A, B, C, D, E, F, T }   A -24 B  S -9 -18  -6 -2 E  C  -19 -30 -11 D -15 -5 -6 -25  -20 T F 

  21. Example Problem S = { } Q = { S, A, B, C, D, E, F, T } Source   A -24 B  S -9 -18  -6 -2 E  C  -19 -30 -11 D -15 -5 -6 -25  -20 T F 

  22. Example Problem S = { S } Q = { A, B, C, D, E, F, T } update   -9 X A -24 B  update S -9 -18  -9 X -6 -2 E  C  -19 -30 -11 D -15 -5 -6 -25  update -20 T F  -15 X

  23. Example Problem S = { S } Q = { A, B, C, D, E, F, T }   -9 X A -24 B  S -9 -18  -9 X -6 -2 E  C  -19 -30 -11 D -15 -5 -6 -25  -20 T F Min energy  -15 X

  24. Example Problem S = { S, F } Q = { A, B, C, D, E, T }   -9 X A - 24 B  S -9 -18  -9 X -6 -2 E  C  -19 -30 -11 D -15 -5 -6 -25  -20 T F  -15 X

  25. Example Problem S = { S, T, A, B, C, D, E, F } Q = { } -9  X  -9 X A -24 B  S -9 -18  -9 X -6 -2 E  -11 X  -15 C -30 X -19 -11 D -15 -5 -6 -25  -15 X -20 T F  -15 X

  26. Dominant-Edge Path Node D is an intermediate byproduct -9  X  -9 X A -24 B  S -9 -18  -9 X -6 -2 E  -11 X  -15 C -30 X -19 -11 D -15 -5 -6 -25  -15 X -20 T F  -15 X

  27. Augmenting Path Path from byproducts to final product -9  X  -9 X A -24 B  S -9 -18  -9 X -6 -2 E  -11 X  -15 C -30 X -19 -11 D -15 -5 -6 -25  -15 X -20 T F  -15 X

  28. Runtime • Dominant Path: O(|V|2 + |E|) • Augmenting Path: O(|V|2 + |E|)

  29. Test Cases • Pentose Fermentation in Zymomonas mobilis (bacterium) • Escherichia coli • Liver Cell

  30. Test Case 1 (Altintas, et al., 2006)

  31. Test Case 2 (Wlaschin, et al., 2006)

  32. Test Case 2 Total elementary flux modes : 33,000 Each DOMINANT-EDGE pathway between an input sugar and ethanol exactly corresponds to the maximal conversion pathway engineered through gene knockouts by Wlaschin et al., 2006

  33. Test Case 3 • Total elementary flux modes : 188

  34. Runtime • Windows Machine • C++ Implementation • 4GB Memory

  35. ∆G Used in Recent Studies • ∆G analysis of a E. Coli resulted in identifying four reactions representing thermodynamic bottlenecks in the production of growth (Henry et al., 2006) • Using thermodynamics-based constraints to enforce the exclusion of thermodynamically infeasibilities when calculating flux distribution (e.g. a flux cannot be positive unless ∆G is negative) (Henry et al., 2007)

  36. Conclusion • An Efficient way of finding a maximum conversion pathway with runtime advantages over enumeration-based approaches • The Dominant-Edge Pathway found is identical, proper subset, or partially overlaps with EFM results

  37. Applications & Future Work • Pathway Analysis has promising applications • Understanding living systems • Efficient in silicoexperimentation • Engineering microorganisms to perform specific synthesis tasks • Future Work • Find an energetically favored pathway (most negative overall Gibbs energy change) • Find alternative pathways with (nearly) equivalent bottleneck or pathway weights

  38. Acknowledgements • Gautham Sridharan (BCE Tufts University)

  39. Questions

  40. Thank You

  41. Questions for Ehsan? • Why use EFM instead of Extreme Pathways? How would your results be different? • How did you calculate the delta_Gs? • How do you ensure that your paths are stoichiometrically balanced in terms of conversion?

More Related