Evolvability and CrossTalk in Chemical Networks

1
Evolvability and Cross-Talkin Chemical Networks

• Chrisantha Fernando
• Jon Rowe
• Systems Biology Centre
• School of Computer Science
• Birmingham University, UK
• ESIGNET Meeting September 2007

2
Aims

• Model evolution and function of cellular networks
• Understand the principles of evolvability in
  cellular networks
• Model cross-talk

3
Simulated Evolution of Protein Networks
Bray and Lay, 1994
4
L
T
Tp
Tp
5
(No Transcript)
6
Conclusions

• The genetic description of proteins used was
  very unevolvable, i.e. brittle.
• Stochastic simulation did not allow futile
  cycles to be modeled efficiently. These are
  essential for information transmission.
• We moved to a more abstract representation of
  chemical networks, inspired by work in Eindhoven.

7
Turing Complete Enzyme Computers
To Appear in European Conference in Artificial
Life 2007 Lisbon.
8
(No Transcript)
9
(No Transcript)
10
Conclusion

• Although there is now an easy way of programming
  serial programs in enzyme controlled systems.
• Implementation in a physical system is not
  trivial!!
• Parallel implementations are possible.
• But how could we get evolvable chemical networks
  in the real world?

11
Chemical Evolution
12
How did metabolism evolve?
13
The Chemoton
Metabolism
Template
Membrane
14
(No Transcript)
15
Catalysis
Autocatalysis

• Molecular autocatalysts are necessary for
• heredity.
• Some have 2o effects that are beneficial
• to the compartment.
• Some energy is required for this memory.

16
Substrate
New autocatalysts arise and integrate into
existing intermediary metabolism Not a reflexive
autocatalytic set!
17
Multiplication Yes Heredity Yes Variability
Macro not micro
18
(No Transcript)
19
Is there a limit to complexity increase? Yes, in
this simple model, the probability of stable
autocatalyst formation decreases!
20
The metabolic equivalent of Szathmarys SCM
21
Conclusions

• A limit to complexity is imposed if chemical
  variability properties cannot be shaped by second
  order selection
• Self-isolation of faulty components (Tan,
  Revilla, Zauner, 2005)
• What is second-order selection?

• Real chemicals embody variability rules as
  (modular) structures.
• Make a chemical description language capable of
  representing chemical equivalence classes
  abstractly, that allows adaptive variability.
• Evolve the system at the compartment level to
  maximize information transmission.

22
Second order selection is selection on the basis
of offspring fitness
A
23
It can act on variability properties
B
24
Evolvability shaped by second order selection?

• Produce a CE-calculus, capable of representing
  the crucial functional properties of small
  molecules that allow them to be structured by
  second order selection to promote evolvability,
  information transmission, and effective search.
• Use Keppa (Vincent Danos, Harvard)

25
European Collaborations Arising

• Eors Szathmary, ThalesNano (Budapest) Guenter
  Von Kiedrowski (Bochum), FP7 Large scale
  application.
• Evolution of Formose cycle combinatorial
  libraries
• Find lipid precursor that reacts with formose
  cycle sugars via phase-transfer autocatalysis
  yielding sugar-lipid conjugates.
• Study the formose cycle using such a precursor
• Study these subsystems under high pressure

26
A New Kind of Cell Signaling using RNAi

• Protein structure to function map is very
  complex.
• A simpler and possibly more evolvable CSN could
  be made from RNA.
• John Matticks work shows the large amount of
  non-translated RNA in cells.
• We published a simulator capable of modeling
  complex populations of interacting RNA molecules
  with simple 2o structures.

27
Bad Cross-Talk Side-Reactions
28
d
a
c
b
29
(No Transcript)
30
Minimal Replicase was a Restriction Ribozyme
David Bartel and Jack Szostak barking up wrong
tree?
31
Conclusions

• The simulator used a simplified model of nucleic
  acid interactions to test hypotheses about how
  autocatalytic RNA could function in the absence
  of protein enzymes.
• Further work will increase the range of secondary
  structures, e.g. hairpins.

32
Bacteria that can learn

• Replicate this experiment
• Is learning epigenetically heritable?
• Are there any associated macro-nuclear
• gene changes? (L. Landweber)

33
Cross-talk does association

• In collaboration with molecular biologists,
  (Prof. Pete Lund, Dr. Lewis Bingle) and Anthony
  Liekens we have designed Hebbian learning
  circuits in plasmids carried by E. coli.

v w.u dwi/dt uiv
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
Peter Dittrich, Thorsten Lenser Christian
Beck Evolver uses Biobrick primitives. It is
a Synthetic Biology Toolbox
40
What to expect?
41
Later.
42
Cell Signaling Network Implementation
43
Conclusions

• Nature paper in prep.
• Grant applications for synthesis in prep.
• Future medical applications.
• Introduces learning concepts to systems biology.

44
Liquid State Machines in Bacteria?
45
Why so Little Lamarckian Inheritance?
ECAL 2007
46
Publications so far
http//www.cs.bham.ac.uk/ctf/
47
Expected Publications

• Nature. Hebbian Learning (in collaboration with
  Eindhoven and Jena).
• Evolution. Second-order selection for
  evolvability.