This is a plenary talk I gave at the 2018 International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems in Cadiz, Spain

1. ‘Biological Apps: Rapidly
Converging Technologies for
Living Information Processing
Natalio Krasnogor
Newcastle University
Twitter: @Nkrasnogor
http://Ico2s.org
IPMU 2018 – Cadiz, Spain

2. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
•Molecular Computation: A DNA based data structure
•Conclusions & Opportunities for IPMU Community
2IPMU 2018 – Cadiz, Spain

3. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
•Molecular Computation: A DNA based data structure
•Conclusions & Opportunities for IPMU Community
3IPMU 2018 – Cadiz, Spain

4. The Spatial Scales Involved
Origami
1 nm = 10-9 m
4IPMU 2018 – Cadiz, Spain

5. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
•Molecular Computation: A DNA based data structure
•Conclusions & Opportunities for IPMU Community
5IPMU 2018 – Cadiz, Spain

6. Living Cells Are Information Processors
LeDuc et al. Towards an in
vivo biologically inspired
nanofactory. Nature (2007)
6IPMU 2018 – Cadiz, Spain

7. The Cell as an Intelligent (Evolved)
Machine
Wikimedia Commons
7IPMU 2018 – Cadiz, Spain

8. Just in Time Hardware Assembly/Disassembly
Gene1 Gene2 Gene3 Genek
Genome
Transcription Factors
Signal2 Signal5Signal1 Signal3 Signal4 Signaln...Environment
8IPMU 2018 – Cadiz, Spain

9. 9IPMU 2018 – Cadiz, Spain

10. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
•Molecular Computation: A DNA based data structure
•Conclusions & Opportunities for IPMU Community
10IPMU 2018 – Cadiz, Spain

11. Agent-based modelling platform for multicellular
systems, focusing on bacterial populations such as
biofilms.
Simulating a spatially explicit physical world with
chemical and biological processes.
Allows modelling of feedback between cell
genetics and population behaviour.
Runs on laptop but can scale to millions of cells on
HPC (multi-threaded & multi-CPU).
Flexible modelling environment for rapid
prototyping.
Multi-Scale Simulator
11IPMU 2018 – Cadiz, Spain

12. Design
a
Specify species behaviour
Set positions of cells
Set environmental properties
Simulation
a
3D physical simulations
Multi-thread and Multi-CPU
Analysis
a
Probe model properties
Run virtual lab machines
Take pictures/video
12IPMU 2018 – Cadiz, Spain

13. Modular platform architecture and model design
Physics
- 3D geometries, forces
- Boundary conditions
Chemistry
- diffusion
- reactions, degradation
Biology (cell processses/behaviours)
- Motility (run & tumble, chemotaxis)
- Metabolism, gene regulation
- Membrane transport
- Surface-mediated interactions (receptor-adhesin)
- Conjugation
Expressing cell dynamics
- Ordinary differential equations, Gillespie stochastic simulations,
boolean networks, SBML submodels
- Conditional actions (eg. Cell age > T ? grow_flagellar)
13IPMU 2018 – Cadiz, Spain

14. Colonies of rod-shaped cells (microscopy images)
colony in scarce substrate
Population Structure Modeling
14IPMU 2018 – Cadiz, Spain

15. Friction
15IPMU 2018 – Cadiz, Spain

16. Pattern formation
colony in scarce substrate
Signal Propagation Modeling
In collaboration with:
Francisco Campero-Romero, Plant
Development Unit, Institute for Plant
Biochemistry and Photosynthesis,
Consejo Superior de
Investigaciones Científicas,
Universidadde Sevilla, Seville,
Spain
16IPMU 2018 – Cadiz, Spain

17. Signal Propagation
17IPMU 2018 – Cadiz, Spain

18. colony in scarce substrate
In collaboration with:
Francisco Campero-Romero, Plant Development Unit, Institute for Plant Biochemistry and
Photosynthesis, Consejo Superior de Investigaciones Científicas, Universidadde Sevilla, Seville,
Spain
18IPMU 2018 – Cadiz, Spain

19. Signal Propagation
T = 5  T = 20  T = 40
Simulation showing pulse propagation across the colony (cells are purple, pulse green)
Pulse velocity was measured for different
values of the signal diffusion coefficient D
and signal degradation coefficient K
We find that the pulse velocity is highest
when the diffusion rate and degradation
rate are low, and slowest when both are
high.
19IPMU 2018 – Cadiz, Spain

20. Pattern formation
Turing-like pattern formation system.
20IPMU 2018 – Cadiz, Spain

21. Pattern formation
Visualisations of colonys showing flourescent protein expression F1 (green) and F2
(red), for different signal diffusion coefficient D and signal degradation coefficient K. S1
and S2 are set to have equal D and K values. The system is induced by pipetting S1 at
the center.
K high
D high21IPMU 2018 – Cadiz, Spain

22. Turing Patterns in Bacterial Colonies
22IPMU 2018 – Cadiz, Spain

23. In Silico & In Vivo
23IPMU 2018 – Cadiz, Spain

24. Conjugation rate r:
r = 0.1 r = 0.01 r = 0.001
BioProcessors exchange & adapt hardware/instructions + reproduce!
 fuzzy signal processing and pattern formation
2D biofilm growth
under scarce
substrate
Different Growth Rate
24IPMU 2018 – Cadiz, Spain

25. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
•Molecular Computation: A DNA based data structure
•Conclusions & Opportunities for IPMU Community
25IPMU 2018 – Cadiz, Spain

26. Combinatorial DNA Synthesis on your Desktop
IWBDA 2016 - Newcastle Upon Tyne
Parts
Library Targets
Operators
Planer Assembly Plan
Instrumentation
Programable Order
Polymerization (POP)
Microfluidics Combinatorial
Assembly of DNA (M-CAD)
Microfluidics In Vitro
Cloning (MIC)
Key challenge is to enable precise
design, editing and manufacturing
of combinatorial DNA libraries at
your desk.
CAD
CAM
26IPMU 2018 – Cadiz, Spain

27. A Programming Language for Sequences:
DNALD (DNA Library Design)
A specification language that
produces a set of target DNA
sequences as a function of
operations on a set of inputs
To maximise impact the specification process must be:
• user friendly and debuggable
• but expressively powerful enough to:
– define non-trivial combinatorial constructs
– communicate degrees of freedom
IWBDA 2016 - Newcastle Upon Tyne
27IPMU 2018 – Cadiz, Spain

28. DNA Library Designer with DNALD
IWBDA 2016 - Newcastle Upon Tyne
28IPMU 2018 – Cadiz, Spain

29. Background Validation
evaluation
constraints
syntax
errors
error
navigation
errors
marked
29IPMU 2018 – Cadiz, Spain

30. Versioning
duplicate each or
every change
highlights insertions
and deletions
30IPMU 2018 – Cadiz, Spain

31. And Paired Visualisations
 Emphasizes reuse with shared nodes and provides
indication of library's combinatorial degree
 Every path from 5' to 3' is an output
Graphical Representation of
Complex DNA Libraries
31IPMU 2018 – Cadiz, Spain

32. From the DNA Library to the
Synthesis Plan
 When O={+} & P=unrestricted 
Planning problem
 Related computational problem
“Bounded-depth min-cost string
production (BDMSP)” is NP-hard
and APX-hard by reduction from
vertex cover
32IPMU 2018 – Cadiz, Spain

33. Combinatorial DNA Synthesis on your Desktop
Parts
Library Targets
Operators
Planer Assembly Plan
Instrument Instructions
Programable Order
Polymerization (POP)
Microfluidics Combinatorial
Assembly of DNA (M-CAD)
Microfluidics In Vitro
Cloning (MIC)
CAD
CAM
33IPMU 2018 – Cadiz, Spain

34. Programmable Liquid Handling
34IPMU 2018 – Cadiz, Spain

35. 35IPMU 2018 – Cadiz, Spain

36. Stem Cell Reprogramming (UKB)
Frank Edenhofer
36IPMU 2018 – Cadiz, Spain

37. Synthetic Circuits Rewriting (UEVE)
François Képès
37IPMU 2018 – Cadiz, Spain

38. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
• Molecular Computation: A DNA based data
structure
•Conclusions & Opportunities for IPMU Community
38IPMU 2018 – Cadiz, Spain

39. • DNA has been used to store data.
• DNA can store very large amount of data.
• DNA is a durable, efficient and cheap digital substrate.
• DNA data structures for information processing.
• Biological data with programmatic API
Fellermann H., Lopiccolo A., Kozyra J., Krasnogor N. (2016) In Vitro
Implementation of a Stack Data Structure Based on DNA Strand
Displacement. In: Amos M., CONDON A. (eds) Unconventional Computation
and Natural Computation. UCNC 2016. Lecture Notes in Computer Science,
vol 9726. Springer, Cham
39IPMU 2018 – Cadiz, Spain

40. A stack is a data register, two operations: push and pop. You can
push values on the stack, and pop them from the stack. This
happens in LIFO (last in, first out) order.
SIGNAL (X)
SIGNAL (Y)
SIGNAL (Y)
SIGNAL (X)
PUSHING POPPING
Last in First out
Stack Data Structure
40IPMU 2018 – Cadiz, Spain

41. Image by Ouldridge et al.
The rate constant of the strand-displacement reaction varies over a factor of 106,
from 1 M−1 s−1 to 6 × 106 M−1 s−1.
DNA hybridization and DNA strand displacement
41IPMU 2018 – Cadiz, Spain

42. DNA “Bricks”
Smallest brick: 22nt Largest brick: 137nt
42IPMU 2018 – Cadiz, Spain

43. Stack Recorder - Single Molecule Operation
start
Last In
First Out
Recording 4 signals on stack
in order X(1), Y(1), X(2), Y(2)
43IPMU 2018 – Cadiz, Spain

44. Stack Recorder - Single Molecule Operation
start
push
Last In
First Out
44IPMU 2018 – Cadiz, Spain

45. Stack Recorder - Single Molecule Operation
Last In
First Out
45IPMU 2018 – Cadiz, Spain

46. Stack Recorder - Single Molecule Operation
signal X(1)
X(1)
Last In
First Out
46IPMU 2018 – Cadiz, Spain

47. Stack Recorder - Single Molecule Operation
X(1)
Last In
First Out
X(1)
47IPMU 2018 – Cadiz, Spain

48. Stack Recorder - Single Molecule Operation
push
X(1)
Last In
First Out
X(1)
48IPMU 2018 – Cadiz, Spain

49. Stack Recorder - Single Molecule Operation
X(1)
Last In
First Out
X(1)
49IPMU 2018 – Cadiz, Spain

50. Stack Recorder - Single Molecule Operation
signal Y(1)
X(1)
Last In
First Out
X(1)
Y(1)
50IPMU 2018 – Cadiz, Spain

51. Stack Recorder - Single Molecule Operation
X(1)
Y(1)
Last In
First Out
X(1)
Y(1)
51IPMU 2018 – Cadiz, Spain

52. Stack Recorder - Single Molecule Operation
After 4 signals pushed to stack:
X(1)
Y(1)
X(2)
Y(2)
Last In
First OutX(1)
Y(1)
X(2)
Y(2)
52IPMU 2018 – Cadiz, Spain

53. DNA Sequence Optimisation
Domains on the DNA bricks
had their nucleotide sequences optimised
so that these multiple objectives were satisfied:
Single DNA strands folded into the
correct local topology
Pairs of DNA strands co-folded into the
correct stack topology
Desired reactions resulted in irreversible transformations
Undesired reactions were minimised
Jerzy Kozyra, Harold Fellermann, Ben Shirt-Ediss, Annunziata
Lopiccolo, and Natalio Krasnogor. 2017. Optimizing nucleic
acid sequences for a molecular data recorder. Proceedings
of the Genetic and Evolutionary Computation Conference
(GECCO '17). ACM, New York, NY, USA, 1145-1152. DOI:
https://doi.org/10.1145/3071178.3071345
53IPMU 2018 – Cadiz, Spain

54. S + P + X
Desired Multi-Molecule Scenario
All stack complexes in solution would have identical state
54IPMU 2018 – Cadiz, Spain

55. Actual Multi-Molecule Scenario
S + P + X
— Complexes can have several isoforms
— Unintended side reactions take place
— DNA complexes have finite diffusion and reaction rates
— Finite wait times mean that chemistry is still under kinetic
control
55IPMU 2018 – Cadiz, Spain

56. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
Single Tube Experimental Results
56IPMU 2018 – Cadiz, Spain

57. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SP
57IPMU 2018 – Cadiz, Spain

58. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPX
1 signal
58IPMU 2018 – Cadiz, Spain

59. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXP
59IPMU 2018 – Cadiz, Spain

60. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPX
2 signals
60IPMU 2018 – Cadiz, Spain

61. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXP
61IPMU 2018 – Cadiz, Spain

62. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXPX
3 signals
62IPMU 2018 – Cadiz, Spain

63. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXPXP
63IPMU 2018 – Cadiz, Spain

64. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXPXPX
4 signals
64IPMU 2018 – Cadiz, Spain

65. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXPXPXP
…and so on
65IPMU 2018 – Cadiz, Spain

66. S
P
S
P
X
S
P
X
P
S
P
X
P
X
S
P
X
P
X
P
S
P
X
P
X
P
X
S
P
X
P
X
P
X
P
S
P
X
P
X
P
X
P
X
SPXPXPXPXP
…and so on
66IPMU 2018 – Cadiz, Spain
Fuzzy Molecular Data Structure

67. Outline
• Multi-Scale Computation in Nature
• Living Cells are Information Processors
• Simulations as virtual machines for cells
• How to program cells
• Molecular Computation: A DNA based data
structure
• Conclusions & Opportunities for IPMU Community
67IPMU 2018 – Cadiz, Spain

68. Living cells are stochastic, asynchronous & highly parallel
bio-processors & constructors that adapt and generate
their own hardware on-demand
Information processing is organised via interconnected
networks (genes, signaling, metabolic, etc)
DNA/RNA computation even more parallel, processing
networks more dense
In principle fully definable but in practice great sources
of variability and uncertainty
68IPMU 2018 – Cadiz, Spain

69. The Smaller You Go The Dumber
The Processors
Computational Design of DNA/RNA
Origami
69IPMU 2018 – Cadiz, Spain

70. The Smaller You Go The Dumber
The Processors
Tetra-Pyridyl Porphyrin (TPyP) on Au(111)
Structural unit
to functionalise
(~ 1nm)
With G. Terrazas, P.
Moriarty and N.
Chapness
70IPMU 2018 – Cadiz, Spain

71. Backbone
Self-assembly
counting process
• Blue porphyrin-tiles act as counters1,2 “seeded”
via red porphyrin-tiles
• Backbones are spatial limits controlling blue-
porphyrin-tiles assembly
1 Q. Cheng et al. Optimal self-assembly of counters at temperature two. In Foundations of Nanosciense, 2004.
2 P. Moisset. Computer aided search for optimal self-assembly systems. In N. Krasnogor et al. (Eds.), Systems Self-Assembly Multidisciplinary Snapshots, 2008.
m1
m2
Embedded Discrete Process of Computation (I)
Backbone
Es = 0.50
E11 = 1.00
E22 = 0.20
E12 = 0.20
Es = 0.60
E11 = 0.40
E22 = 0.20
E12 = 0.10
71IPMU 2018 – Cadiz, Spain

72. Embedded Discrete Process of Computation (II)
Checkers pattern
(spatial interactions)
• Highly ordered self-assembled
structure
• Spontaneous internal arrangements
• Globally complex shape with locally
simple organisation
λ (y)
λ (y)
(x)
(ε) (ε)
(x)
q1
q2
ε, x, y Є [0, 1]
ε + x + y = 1
x >> ε >> y
Computed by a finite
state machine-like
process
ε: probability of mistaking symbol
λ: new diagonal begins
72IPMU 2018 – Cadiz, Spain

73. Es = 0.50 E11 = E22 = 0.10 E12 = 0.40
Es = 0.50 E11 = E22 = 0.10 E12 = 0.30
Es = 0.50 E11 = E22 = 0.30 E12 = 0.40
Es = 0.50 E11 = E22 = E12 = 0.30
Differently programmed spatial
interactions generate:
• micro level features
(order/disorder)
• macro level features
(regular/irregular shape)
73IPMU 2018 – Cadiz, Spain

74. Multiscale Computation in Nature
• A Research Programme Programmable algorithmic entry to
the vast world of nanoscale physical,
chemical & biological systems and
processes
ComputerScience
Information & Algorithms
Embedded behavior
Robustness
Uncertainty
Complexity
Tradeoffs
How does “The Logistics of Small Things” look like?
How does “The Decision Making in/with Small Things” take place?
How is “Uncertainty Handled by Small Things” ?
74IPMU 2018 – Cadiz, Spain

75. Uncertainty & Fuzziness
Everywhere
• At the component level
• At the interactions level
• At the orchestration of interactions level
• Known and Unknown sources of variability & errors
• Nature evolved counters, clocks and consensus mechanisms to give
robustness to its systems
• Need for new ways of looking at information processing in nanobio:
– We spend too much effort pursuing crisp, digital results
– Maybe a fuzzy approach more realistic and pragmatic
– Great opportunity for the Information Processing and Management of
Uncertainty community
75IPMU 2018 – Cadiz, Spain
• Fuzzy Signal Processing
• Fuzzy Pattern Formation
• Fuzzy JIT Hardware
Assembly
/ Disasembly
• Fuzzy Data Structures

76. Thank you
• IPMU organisers
• UK’s EPSRC for funding
• The brilliant PhD students and postdocs
76IPMU 2018 – Cadiz, Spain