JSPS Project on Molecular Computing
(presentation by Masami Hagiya)
• funded by Japan Society for Promotion of Science
• Research for the Future Program
– biocomputing field chaired by Prof. Anzai
•
•
•
•
•
molecular computing
artificial cell (chemical IC)
evolutionary computation
input/output for molecular computing
complex systems
• October 1996 - March 2001
– 60M-80M yen per year
JSPS Project on Molecular Computing
• project leader - Masami Hagiya (Computer Science)
• members
–
–
–
–
–
–
–
Takashi Yokomori (Computer Science)
Akira Suyama (Biophysics)
Yuzuru Husimi (Biophysics)
Kensaku Sakamoto (Biochemistry)
Shigeyuki Yokoyama (Biochemistry)
Masayuki Yamamura (Computer Science)
Masanori Arita (Genome Informatics)
Some Achievements
• Kensaku Sakamoto, Shigeyuki Yokoyama, and Masami Hagiya
– Whiplash PCR
– Hairpin Engine
• Masami Hagiya
– VNA --- a simulator for DNA reactions
• Takashi Yokomori
– Formal Models of DNA Computing
• Akira Suyama
– Solid-based DNA Computing
– Dynamic Programming DNA Computers
• Yuzuru Husimi
– Evolutionary Reactor Based on 3SR
• Masayuki Yamamura
– Aqueous Computing (with Tom Head)
DNA Computing
• Adleman-Lipton Paradigm ＝ massive parallelism
⇒ combinatorial
– random generation by assembly
– selection by molecular biology experiments optimization
large size ⇒ increase yield and decrease error
• one-pot reaction ⇒
autonomous computation ・ self-organization
– DNA tiles by Winfree
– DNA automata by Hagiya et al.
– DNA boolean circuits by Ogihara and Ray
• theoretical analyses
– complexity
– formal languages
Brute Force v.s. Dynamic Programming
DNA Computers
Brute Force
Dynamic Programming
Large pool size
Low reaction rate
small pool size
High reaction rate
Generating
WHOLE
solution space
AT ONCE
Generating
PARTIAL
solution space
STEP BY STEP
Selection
Selection
Solution
Solution
3-CNF SAT Solution on DP DNA Computer
Problem : 4 variables, 10 clauses
( x1  x 2  x3 )  ( x1  x 2  x3 ) 
(x1  x 2  x3 )  (x1  x 2  x3 ) 
( x1  x3  x 4 )  (x1  x 2  x 4 ) 
(x1  x3  x 4 )  ( x 2  x3  x 4 ) 
( x 2  x 3  x 4 )  ( x 2  x 3  x 4 )
Solution :
YES
{ X 1T X 2T X 3F X 4F }
Basic Operations for
Dynamic Programming DNA Computers
get (T, +s), get (T, -s)
get DNA molecules with a subsequence s (without s)
in a tube T
append (T, s, e)
append a subsequence s at the end of DNA molecules
with a splint e in a tube T
merge (T, T1, T2, …, Tn)
merge DNA molecules in tubes T1, T2, …, Tn into a
tube T
amplify(T, T1, T2, …, Tn)
amplify DNA molecules in a tube T and divide them
into tubes T1, T2, …, and Tn
detect(T)
detect DNA molecule in a tube T
Implementation of Basic Operations
get (T, +s), get (T, -s)
annealing
and
ligation
annealing
s
annealing
Taq DNA ligase
T
s
s
s
amplify (T, T1, T2, …Tn)
append (T, s, e)
e
PCR
immobilization
and
cold wash
immobilization
s
s
s
e
immobilization
and
cold wash
s
hot wash
cold wash
s
get (T, -s)
hot wash
s
hot wash
and
divide
T1, T2, …Tn
get (T, +s)
DP algorithm for 3CNF-SAT
on DNA Computers
procedure dna 3 sat (u1 , v1 , w1 , ... , um , vm , wm )
begin
T2T  { X 1T X 2T , X 1F X 2T }; input (T2T ); T 'T2  { X 1T X 2T , X 1F X 2T }; input (T 'T2 );
T  { X X , X X }; input (T ); T '  { X X , X X }; input (T ' );
F
2
T
1
F
2
F
1
F
2
F
2
F
2
T
1
F
2
F
1
F
2
for k  3 to n do
merge (TwT , TkT1 , TkF1 ); merge (TwF , T 'Tk 1 , T 'kF1 );
for j  1 to m do
input (T );
TuT  get (T ,  X uT ); T 'uF  get (T ,  X uT );
TuF  get (T 'uF ,  X uF );
TvT  get (TuF ,  X vT );
merge (T T , TuT , TvT );
if w j ' xk ' then
T  getuvsat (T , u j , v j );
F
w
F
2
procedure getuvsat (T , u , v)
begin
F
w
output (T T );
end
end
if w j ' xk ' then
TwT  getuvsat (TwT , u j , v j );
end
Number of operations
end
T T  append (TwT , X kT , X kT/1F X kT ); T F  append (TwF , X kF , X kT/1F X kF );
amplify (T T , TkT , T 'Tk ); amplify (T F , TkF , T 'kF );
end
output (detect (Tn ));
end
n  (2  merge  2  append
 2  amplify)

3m  get  m  merge
An Amount of DNA for Computation
Brute Force v.s. Dynamic Programming
100 variable 3-CNF SAT
Adleman-Lipton’s
Brute Force
DNA Computers
2x1012 g of dsDNA
1x1012 g of ssDNA
Dynamic Programming
DNA Computers
2x10-3 g of dsDNA
(1x10-3 g of ssDNA)
Autonomous Computation
Self-Organization
• computational power of assembly (Winfree)
– string
– tree
– tile
― regular
― context-free
― universal
• autonomous state transition (Hagiya et al.)
– Whiplash PCR
• applications ― not only combinatorial optimization, but also
– nanotechnology
– genetic analysis
Whiplash PCR
x
b
x
B
B
a
A
x
C
Whiplash PCR
B
x
B
A
x
C
Whiplash PCR
a
x
B
A
x
C
B
x
Whiplash PCR
a
c
x
B
A
x
C
b
B
x
Encoding
VNA: Simulator for Virtual DNA
• a concrete computational model based on
assembly and state transition
• abstract, but sufficiently physical
bridging gaps between abstract models and real reactions
molecule ― hybrid of virtual strands
• reactions
–
–
–
–
–
–
abcd
||
CDEF
hybridization
denaturation
restriction
ligation
self-hybridization
extension
assembly
decomposition
decomposition
state transition
state transition
state transition
VNA (cont.)
• purposes
– verify feasibility of algorithms for DNA computing
– verify validity of molecular biology experiments
(e.g., PCR experiments)
– parameter fitting in molecular biology experiments
• examples
– Ogihara and Ray’s computation of boolean circuits
– Winfree’s construction of double-crossover units
– PCR experiments
• implementation
– Java ⇒ executable as an applet
VNA (cont.)
• methods
– combinatorial enumeration
– continuous simulation (diff. eq.)
unify
• avoiding combinatorial explosion
contributions in simulation technology
– threshold
– stochastic
• parameter fitting by GA
– optimizing amplification in PCR experiments
Goals of Molecular Computing
― concluding remark
• analysis of computational power of biomolecules
– understanding life
• origin of life
• wet artificial life
– engineering applications
• combinatorial optimization
• nanotechnology
• genetic analysis
• new computational models ・ simulation technology
digital ⇒ analog ⇒ physical