automated)design)of)synthe3c) biology)feedbackcircuits) · • dead code elimination •...

Automated Design of Synthe3c Biology Feedback Circuits

Jacob Beal, Aaron Adler

IBE Conference March, 2012

Work partially sponsored by DARPA; the views and conclusions contained in this document are those of the authors and not DARPA or the U.S. Government. 1

Vision: WYSIWYG Synthetic Biology

Bioengineering should be like document preparation:

2

Why is this important?

•  Breaking the complexity barrier:

•  Multiplication of research impact •  Reduction of barriers to entry

*Sampling of systems in publica?ons with experimental circuits

207

2,100 2,700

7,500 14,600

32,000

583,000 1,080,000

100

1,000

10,000

100,000

1,000,000

1975 1980 1985 1990 1995 2000 2005 2010

Length in base pa

irs

Year

DNA synthesis Circuit size ?

3

[Purnick & Weiss, ‘09]

Why a tool-chain?

Organism Level Descrip3on

Cells

This gap is too big to cross with a single method!

4

The TASBE architecture:


Abstract Gene3c Regulatory Network

DNA Parts Sequence

Assembly Instruc3ons

Cells

High level simulator

Coarse chemical simulator

Testing

High Level Descrip3on If detect explosives: emit signal If signal > threshold: glow red

Detailed chemical simulator

Modular architecture also open for flexible choice of organisms, protocols, methods, …

5

Ron Weiss

Douglas Densmore

Collaborators:

A Tool-Chain Example

(yellow (not (cyan (Dox))))!

Dox cyan not yellow

rtTA CFP B EYFP

Dox

pHef1a rtTA pTre CFP pTre LacI pHef1a-‐LacO1Oid

EYFP mirff4 4xff4

6

Today’s focus: BioCompiler


Abstract Gene3c Regulatory Network

DNA Parts Sequence

Assembly Instruc3ons

Cells

High level simulator

Coarse chemical simulator

Testing

High Level Descrip3on If detect explosives: emit signal If signal > threshold: glow red

Detailed chemical simulator

Compila3on & Op3miza3on

7

Transcriptional Logic

8

Motif-Based Compilation

•  High-level primitives map to GRN design motifs –  e.g. logical operators:

(primitive not (boolean) boolean :grn-motif ((P high R- arg0 value T)))

arg0 value

9

  High-level primitives map to GRN design motifs   e.g. logical operators, actuators:

(primitive green (boolean) boolean :side-effect :type-constraints ((= value arg0)) :grn-motif ((P R+ arg0 GFP|arg0 value T)))

GFP value arg0


10

  High-level primitives map to GRN design motifs   e.g. logical operators, actuators, sensors:

(primitive IPTG () boolean :grn-motif ((P high LacI|boolean T) (RXN (IPTG|boolean) represses LacI) (P high R- LacI value T)))

value LacI

IPTG


11

  Functional program gives dataflow computation:

(green (not (IPTG)))

IPTG not green


12

  Operators translated to motifs:

IPTG not green

LacI

A IPTG

B

GFP outputs outputs outputs arg0 arg0

LacI A

IPTG

B GFP


13

LacI A

IPTG

B GFP

LacI A

IPTG

B GFP

Copy Propaga<on

LacI A

IPTG

GFP

Dead Code Elimina<on

LacI A

IPTG

GFP

Dead Code Elimina<on

Optimization

14

GRN Optimization Library

•  Copy Propagation •  Dead Code Elimination •  Double-Negative Elimination •  Constant Eliminator •  CSE: Output Consolidation •  CSE: Merge Duplicate Inputs •  NOR Compression

15


•  Copy Propagation •  Dead Code Elimination •  Double-Negative Elimination •  Constant Eliminator •  CSE: Output Consolidation •  CSE: Input Merge •  NOR Compression

16

A

B

A

B



17

A



18

A

B

C

C



19

A

B

B



GRN-specific 20

A

B

C

A

A

C



GRN-specific 21

B C

D

A

C D A



GRN-specific 22

C D

A

A

D

A

A

(def sr-latch (s r) (letfed+ ((o boolean (not (or r o-bar))) (o-bar boolean (not (or s o)))) o))

(green (sr-latch (aTc) (IPTG)))

Feedback System: SR-Latch

23

IPTG not green

aTc not or

or



Feedback System: SR-Latch

LacI B

IPTG

I

G

I F

GFP

D E1 E2

A

aTc J

H C

J TetR

Unop<mized: 15 func<onal units, 13 transcrip<on factors 24



Optimization of Complex Designs

LacI B

IPTG

Unop<mized: 15 func<onal units, 13 transcrip<on factors

I

G

I F

GFP

D E1 E2

A

aTc J

H C

J TetR Copy Propaga<on 25

E1




LacI

B

IPTG


I

G

I F

GFP

D

A

aTc J

H C

J

TetR

Common Subexp. Elim. 26

E1




LacI

B

IPTG


I

G

I F

GFP

D

A

aTc H J C

J

TetR

NOR Compression H

27

E1




LacI

IPTG


I

G

I F

GFP

aTc H

TetR

Dead Code Elimina<on H

28

E1




LacI

IPTG


I

G

I F

GFP

aTc H

TetR

Copy Propaga<on H

29

H E1




LacI

IPTG


I G

I F

GFP

aTc

TetR

Common Subexp. Elim. H

30

H




LacI

IPTG


I

I F

GFP

aTc

TetR

Dead Code Elimina<on H

31

H





I F GFP

H

LacI

IPTG

TetR aTc

I

32

H





F GFP

H

LacI

IPTG

TetR aTc

F

NOR Compression

I

I

33

H





F GFP

H

LacI

IPTG

TetR aTc

F

Dead Code Elimina<on 34

GFP


LacI F

IPTG

TetR

H

aTc

F

Final Op<mized: 5 func<onal units 4 transcrip<on factors




H

35

Compilation & Optimization Results:

•  Automated GRN design for arbitrary boolean logic and feedback systems –  Verification in ODE simulation

•  Optimization competitive with human experts: –  Test systems have 25% to 71% complexity reduction –  Optimized systems are often homologous with hand

designed networks •  Routine use within team

36

Onward Through the Tool-Chain…

(yellow (not (cyan (Dox))))!

Dox cyan not yellow

rtTA CFP B EYFP

Dox

pHef1a rtTA pTre CFP pTre LacI pHef1a-‐LacO1Oid

EYFP mirff4 4xff4

37

Contributions

•  Boolean & feedback logic circuits can be expressed in high-level language.

•  GRN-specific optimizations allow dramatic optimization of circuit size.

•  Automatically generated circuits are competitive with hand-designs created by experts.

Next steps: automatic verification, numerical computation, intercellular communication

38

automated)design)of)synthe3c) biology)feedbackcircuits) · • dead code elimination •...

Documents