Topics in Design and Complexity. Herbert M Sauro Bioengineering University of Washington

Transcription

1 Topics in Design and Complexity Herbert M Sauro Bioengineering University of Washington 1

2 Quotes We are still like the Wright brothers, putting pieces of wood and paper together Luis Serrano The field has had its hype phase. Now it needs to deliver. Martin Fussenegger 2

3 Unique Aspects of Synthetic Biology as an Engineering Discipline 1. Liquid technology makes every component unique 2. It evolves makes it unpredictable over the long term 3

4 Complexity in Synthetic Biology 4

5 Complexity in Synthetic Biology 1. Parts are poorly characterized. Since the 80s the tradition of characterizing biological components has been in decline, instead replaced with high-throughput studies. 5

6 Complexity in Synthetic Biology 2. The in vivo context is not well understood. What s the relationship between the characteristics of a component in vitro (the few we do know) compared to the in vivo situation? 6

7 Complexity in Synthetic Biology 3. The context changes most of the time. Context changes even when we think we have a constant and controlled environment. 7

8 Complexity in Synthetic Biology 4. During experiments there is relentless selection and adaptation. 8

9 Complexity in Synthetic Biology 5. Our ability to make quantitative measurements is crude but getting better. 9

10 Complexity in Synthetic Biology 6. We rarely report exactly what we do in an experiment In particular the exact DNA sequence if often omitted from a publication. 10

11 The need for Basic Engineering Research in Synthetic Biology What we lack in synthetic biology is basic knowledge. We are all under pressure today to generated wow results and as inspiring as many of these projects are there is still a great need for research to fill in the gaps and lay a more permanent foundation for synthetic biology. 11

12 Some Real Complexity Linear Pathways Largely understandable without the need to do simulations. Includes branched systems Negative Feedback Positive Feedback Requires some simulations Requires a considerable Degree of simulations to understand Double Positive Feedback: Chaotic Behavior 12

13 Some Basic Experimental Observations 13

14 Short Term Experiment 10^-4 M IPTG 10^-4.5 M IPTG stationary phase In Log phase the properties of the simple circuit are Changing (M9 medium) exponential phase 1) Dynamics of plasmid replication 2) Compensatory effects such as adaptation 3) Changes in metabolic state. 4) Morphological changes In the cells. Back-Dilutions 14

15 Same as Before but Longer Term Experiment 10^-4 M IPTG 10^-4.5 M IPTG exponential phase stationary phase Culture grown for 16hr prior to experiment. Initial culture at very low OD to ensure log growth over the 16hr. Conclusion: Culture eventually reaches steady state. Back-Dilutions 15

16 Effect of Media: Metabolic Switching 16

17 Evolutionary Experiments 17

18 Improving mutational robustness in genetic circuits Canton et al., Nature Biotech

19 T9002 circuit RBS (Ribosome Binding Site) Transcriptional Terminator Promoter Coding sequence BioBrick part Scar sequence Canton et al., Nature Biotech

20 T9002 circuit loses function over evolutionary time due to deletion between homologous terminators Canton et al., Nature Biotech., 2008 Sleight et al., Journal of Biological Engineering, 2010, 4:12 20

21 Circuit re-engineering 1. Can mutational robustness be increased by re-engineering the second terminator? a. Sequence similarity between first and second terminators b. Orientation of the second terminator relative to the first terminator 2. How predictable are mutations in re-engineered circuits? Six circuits were re-engineered and named T9002-A through F. Nine independent populations were evolved for each circuit via serial transfer. Sleight et al., Journal of Biological Engineering, 2010, 4:12 21

22 Evolutionary stability dynamics of re-engineered T9002 circuits Mutation rate Expression High High Med Low Low Med High Low Sleight et al., Journal of Biological Engineering, 2010, 4:12 22

23 Most common mutations in re-engineered T9002 circuits Sleight et al., Journal of Biological Engineering, 2010, 4:12 23

24 Mutations in re-engineered T9002 circuits (500 experiments) T9002 (129**) T9002-A (129 RC) T9002-B (129) T9002-C (41) T9002-D (41 RC) T9002-E (0) T9002-F (0) A* A A A A A A A A A A A A A A A B B A A A A A A A A A A A A A A A A A A A A B A A A A B A A A A A A B C A A A A B A A B A B C 61.9% 12.7% 9.5% 7.9% 7.9% Deletion between homologous sequences (terminators) Deletion between homologous sequences (scars) Point mutation Small insertion or deletion IS element insertion * Letter indicates the uniqueness of a mutation for a particular mutation type ** Sequence similarity between first and second terminators (RC = reverse complement) Sleight et al., Journal of Biological Engineering, 2010, 4:12 24

25 Effect of Expression Level on Evolutionary Robustness 154 Experiments 25

26 Conclusions Promoter mutations (point mutations, deletions) are selected far more than any other effect. Circuits with repeated parts almost always have deletions between these parts, whereas mutations are more random in circuits without repeated parts On average, evolutionary half-life exponentially decreases with increasing expression levels Not all mutations were in the circuit, some occurred in the genome which at least in one case changed the replication rate of the plasmids. Final Note: The use of Kanamycin to protect the circuits did not work. The presence of the Kanamycin gene resulted in other mutations occurring, for example IS bursts due to the additional stress. Sleight et al., Journal of Biological Engineering, 2010, 4:12 26

27 Future directions: Improving mutational robustness of metabolic pathways using directed evolution 27

28 Fan-Out and Signal Propagation 28

29 Gene circuit fan-out A module can be affected by the connections to its downstream. The fan-out is defined as the maximum number of the connections that can be made to a downstream component without affecting the upstream module behavior. How can we measure the fan-out experimentally? Kim and Sauro, Journal of Biological Engineering (2010). 29

30 Visualization of Signal Propagation Track signal propagation in a biochemical system in terms of local response coefficients. Slow S 1 S 2 S 3 S 5 S 4 Fast Perturb S1 by a small amount and measure responses between pairs of species. The propagation of signal is not via single reactions but via reaction pairs. The fundamental dynamical unit is therefore not the single reaction but a pair of reactions. Kim and Sauro, in progress. 30

31 Tracking a Signal Propagation A B C D Movie A B C D A B C D 31

32 SBML Ecosystem SBML Unambiguous Model Exchange Diagrams Databases Journals Semantic Annotations Simulator Comparison and Compliance SEDML: Simulation Experiment Description Language SBGN : Systems Biology Graphical Notation 32

33 SBOL: Synthetic Biology Open Language Overall Aim: To allow researches to electronically exchange designs with round-tripping. To send designs to bio-fabrication centers for assembly. To allow storage of designs in repositories and for publication purposes. 33

34 Synthetic Biology Data Exchange Group Synthetic Biology Open Language (SBOL) SBOL-visual SBOL-semantic ibiosim 34

35 Part subclassof Vector Backbone Sequence Annotation Physical DNA Plasmid subclassof Sample type SS002 type Cell type cell UW002 strain MG1655 dna puw4510 type type vector psb1a2 format BBF RFC 10 type Assembly Standard 1-80 B0010 annotation feature type Terminator insert B annotation feature BioBrick Scar type BioBrick Scar B0012 annotation feature type Terminator type subclassof subclassof Sequence Feature subclassof Figure 1. Diagram of the SBOL Semantic structure, illustrated with a set of information about a synthetic biology construct. a. A simplified Class (black rectangles) hierarchy (black open faced arrows) describes types (colored open faced arrows )of Individual data elements (yellow rounded rectangles) and the composition relationships between them (closed faced arrows). The example can be read as: Sample (pink) SS002 contains UW002 cells (dark green) of the MG1655 E. coli strain, which contain a plasmid (purple) puw4510, which is composed of an parts (dark blue) an insert B0015 and vector backbone psb1a2. The psb1a2 vector backbone complies with the Assembly Standard (light blue) BBF RFC 10. The B0015 sequence annotations (green) specify three features (orange), the BioBrick Scar, and the parts (blue and orange indicating multiple inheritance), B0010 and B0012, which serve as transcriptional termination 35 signals. b. Data type Properties used to hold information for each SBOL class follow the colon.

36 UW Efforts: A CAD system for Synthetic Biology called TinkerCell (Windows, Mac and Linux, released under BSD) Contact author for details (dchandran1@gmail.com) 36

37 Computer-Aided Design (CAD) in Synthetic Biology Challenge How to connect different types of information? SBOL is part of the answer in that it allows us to connect parts repositories with data repositories and models. 37

38 Layered and Extensible Software Architecture 38

39 B.subtilis bistable competence network 39

40 Toggle switch on a plasmid (left) 40

41 Eigenvalue analysis of a terminator repressor 41

42 There are multiple ways to model a biological concept 42

43 Many ways to model a biological network 43

44 TinkerCell can automatically generate multiple models from the same conceptual diagram 44

45 Python plugin that simulates and clusters all possible models 45

46 Mapping concepts to models using Semantics possible sub-models for enzyme catalysis conceptual model 46

47 DREAM Competition Dialogue for Reverse Engineering Assessments and Methods. Primarily organized by Gustavo Stolovitzky (IBM) Build a predictive model of a hypothetical pathogen and on the day of the competition and use the model to identify suitable targets. Between the start of the competition and the final competition day, contestants can purchase experimental data to improve their model. Competition start: Mid May,

48 Acknowledgements: TinkerCell Deepak Chandran and Frank Bergmann: Both Graduates Microsoft and more recently AutoDesk Growth Experiments Bryan Bartley: Graduate Evolutionary Robustness Sean Sleight: Postdoc 1 year of NSF support Engineering Theory Kyung Kim: Postdoc NSF: Theory in Biology SBOL Deepak Chandran (Graduate) and Michal Galdziki (Graduate) DREAM Deepak, Kyung, Lucian Smith, Ryan Roper 48

49 Acknowledging our Collaborators The SBOL Team: UW: John Genarri and Dan Cook (Semantics and Ontologies) BioFAB (Berkeley, Stanford) Cesar Rodriguez, Drew Endy, Adam Arkin, Chris Anderson Utah Chris Myers Virginia Tech Jean Peccoud Boston University Douglas Densmore JBEI (DOE) Timothy Ham & Zinovii Dmytriv CRG, Spain Raik Grunberg 49

50 50

51 Standards and Design Software in silico evolution Global parameter sensitivity analysis Repository of Computational Parts 51

52 Long term goals for TinkerCell 52

53 10^-4 M IPTG 10^-4.5 M IPTG 53

54 10^-4 M IPTG 10^-4.5 M IPTG 54

55 55

56 56

57 10^-4 10^-4 M IPTG 10^-4.5 M IPTG stationary phase exponential phase 57

58 Supplemental Slides 58

59 Experiment 1 Experiment 2 10^-4 M IPTG 10^-4.5 M IPTG exponential phase stationary phase 59

60 Each model component contains detailed information This structure allows loading parameters and other information from database(s) 60

61 Improving mutational robustness in genetic circuits AHL LuxR LuxR GFP T9002 circuit Canton et al., Nature Biotech

62 Module Interfaces vs. Electrical Circuits Fan-out By using = this Max. linear number relationship, of P T that fan-out the output can be can efficiently tolerate. estimated. 62

63 Fan-out Estimation Method Experimental proposal: Signals from fluorescence-tagged transcription factors Simulation Results: Autocorrelation function Output response time Response time linearly increases with P T. This linearity significantly reduces the # of experiments required to estimate the fan-out value. Kim and Sauro, Biophysical Journal (2011). 63

64 Evolutionary half-life exponentially decreases with expression level Sleight et al., Journal of Biological Engineering, 2010, 4:12 64

65 CAD should capture the common region Common region: A conceptual view of the biology 65

66 The Idea behind TinkerCell Database(s) of experimental results Semantically described conceptual diagrams Computational models and are facilitated by Plugins 66

67 Topics 1. Basic issues at the experimental level. 2. Some simple effects of evolution on synthetic circuits. 3. Engineering Aspects: fan-out and signal propagation 4. Standards, CAD and simulation techniques. 67

68 Complexity in Synthetic Biology Systems Biology Workbench 68

69 Complexity in Synthetic Biology Systems Biology Workbench 69

70 Complexity in Synthetic Biology.and yet in spite of all these concerns, synthetic biology has been remarkably successful. 70