Members

1
Members

• Andy Koswara
• Albert Hartono
• Budi Suryawijaya
• Evin Hill
• Hariyono Lie
• Michael Ngantung

2
DNA COMPUTING

• Structure of DNA (Albert)
• Operations of DNA Molecules (Albert)
• DNA to Electrical Device (Budi)
• Traveling Salesman Problem (Michael)
• Hamilton Path (Michael)
• Hamilton Path Solution Using DNA (Hariyono)
• DNA Architecture and Misc. (Evin)
• Conclusion and The future of DNA Computing (Andy)

3
The Structure of DNA

• DNA (Deoxyribo Nucleic Acid) consists three
  components a sugar, a phosphate group, and a
  nitrogenous / base group .
• Sugar has 5 carbon atoms.
• Phosphate group is attached to the fifth carbon.
• Base group is attached to the first carbon.
• There is a hydroxyl group (OH) attached to the
  third carbon.

4

• Chemical structure of a nucleotide with thymine
  base.

5

• Four kinds of base groups
• 1. Adenine (A)
• 2. Thymine (T)
• 3. Guanine (G)
• 4. Cytosine (C)
• Nucleotides can link together in two ways
• 1. Phosphodiester bond
• 2. Hydrogen bond
• Phosphodiester bond is much stronger than
  hydrogen bond.

6
Operations on DNA

• Separating and fusing
• 1. Denaturation heating until 85 C up to 95 C.
• 2. Renaturation slowly cooling down.

7

• Lengthening
• By using polymerases enzymes, we are able to add
  nucleotides to an existing DNA molecule.

8

• A DNA primer sequence is bonded by a single
  stranded template.

9

• Polymerases enzymes in action

10
Shortening and Cutting Nucleases enzymes is an
enzyme that degrades DNA. Two kinds of nucleases
enzymes 1. Exonucleases (for shortening) 2.
Endonucleases (for cutting)
11

• Exonucleases III in action

12

• Exonucleases Bal31 in action

13

• Endonuclease in action

14

• Endonuclease Eco RI in action

15

• Endonuclease XmaI in actio

16
Linking Ligases enzymes is used for DNA
linking. The process of DNA linking is called
ligation.
17

• Hydrogen bonding

18

• Ligation

19

• Blunt ligation

20

• Modifying
• Modifying enzymes is enzymes that modify DNA
  molecule by adding and deleting certain chemical
  components for various control of operations on
  DNA.
• Example
• 1. Alkaline phosphatase
• 2. Polynucleotide kinase

21

• Alkaline phosphatase and Polynucleotide kinase

22
Multiplying Multiplying / amplifying DNA molecule
uses PCR (Polymerase Chain Reaction)
technique. Consists three basic steps 1.
Denaturation / separating 2. Priming 3. Extension
23

• A double-stranded DNA

24

• Denaturation

25

• Priming

26

• Extension

27

• Reference
• Gheorghe Paun, Grzegorz Rozenberg, Arto Salomaa
  DNA Computing. New Computing Paradigms. Springer,
  New York c1998.

28
DNA molecule to electrical device.

• A basic approach to build a DNA processor.

29
Assumptions.

• Chemical bonds(in DNA) can act as tunnel
  junctions in the coulomb blockade regime, could
  emit electricity, given a proper coating.
• Has the ability to coat a DNA strand with metal
  in nanometer scale.

30
Example From DNA to SET transistor.

• SET -gt Single Electron Tunneling transistor.
• In the nanometer scale, and would be able to
  operate at room temperature.
• Function used as an electrometer to measure the
  size and predicted the edge strips of 2DES(2
  dimensional electron system)
• Inspired by a paper from E.Ben Jacob, Z.Hermon
  and S.Caspi from Tel Aviv University.

31
The operation begins(1)

• What a normal SET transistor would look like

32
The Operation(II)

• Schematic image with 2 grains in DNA connected by
  P-bond. Dark circle-gtcarbon atoms, white
  circles-gtoxygen atoms.

33
The operation(II cont) The Battle plan

• P-bond -gt tunneling junction.
• H-bonds -gt capacitor.
• The grain itself -gt inductive properties.

34
The operation(II) The justification(I)

• P bond Has 2 ? bonds, 1 ? bond.
• The ? electron can be shared with 2 oxygen,
  resembles an electron in well, put it on the
  lowest level.
• When electron enters, it meet the barrier set by
  energy gap.
• But the gap is narrow and small so the electron
  can walk trough.

35
The Battle plan(II)The justification(II)

• H-bonds Can be the capacitor.
• The proton in the h-bond can screen a net charge
  density on either side, by movement.
• Thus the net charge could be in the side of the
  h-bond.
• The grains Can be the inductive properties.
• Due to the hopping of additional electrons.
• But can be ignored (L Lo is small, consistent
  to the usual SET)

36
The Battle Plan(III)The finishing touch.

• Consist of 2 strands (1 main, 1 gate)
• Connect the end base of the gate strand with a
  complimentary strand.
• Both strands should be metal-coated, except (a)
  the grain in the main strand, which connect to
  the gate strand, the 2 adjacent P-bonds, (b) the
  connective h-bond.
• Connect the main strand with voltage source (V)

37
The finishing touch (cont)

• The end of the gate strand with another voltage
  source (Vg) that acts as gate source.

38
Some tricky stuff

• Before coating, the DNA molecule should be in a
  solution contain enzyme that can only stick with
  the uncoated part.
• After coating, the enzyme is released, and the
  coating will stick.
• The coated end should be able to be connected
  with a voltage source.

39
Reviewing the assumption

• Is it possible that DNA /other organic molecule
  could carry electricity ?

40
Reviewing the assumption(II)

• Does present technology have the means to coat a
  strand of DNA?
• Currently, the center for nano technology and
  Mitsubishi electric receive 4.2 million to
  develop lithography tools below 35 nano meters.

41
Traveling Salesman Problem
42
Problems

• There is a certain traveling salesman that wants
  to minimize the time and money spent traveling
  around.
• In order to do that he must minimize the distance
  he travels between the cities.
• This is an easy for a computer to solve if there
  are only about 10 cities involved but when there
  are 20 cities it becomes completely impossible.

43
Calculation

• The following calculations shows that it is
  impossible to calculate the traveling salesman
  problem for as many as 20 cities.
• In this calculation, we will assume that the
  computer makes 10 calculations per route and that
  the computer performs 100 million calculations
  per second.

44
Calculation (Continued)

• 10 cities 10! Paths 3628800 paths
• (10 calculations / paths)(3628800 paths) / (100
  million calculations / second) .362 seconds to
  perform the calculation
• 20 cities 20! Paths 2.43 x 10 18 paths
• (10 calculations/paths)(2.43 x 10 18 paths) /
  (100 million calculations / second) 2.433 x 10
  11 seconds to perform the calculation.

45
Calculation (Continued)

• As we can see from the calculation, it takes
  about 7,614 years for the computer to calculate
  the answer.
• Since this Traveling Salesman issue is
  problematic, then we approximate the problem with
  Hamiltonian Path.

46
Hamilton Path

• Hamilton Approximation
• Hamiltonian Path Problem
• Hamilton Algorithm
• Combinatorial Explosion

47
Hamiltonian Approximation

• Hamilton made some important observations back in
  the 19th century that can help us today with
  viewing the traveling salesman problem a little
  more practically
• Instead of optimizing or making the shortest trip
  between all of the cities, our salesman only
  wants to visit all of the cities once
• Each of these cities is considered a vertex on a
  graph. Now the problem is to pass through all of
  the vertices once without breaking the path and
  without passing through anyone twice

48
Hamiltonian Path Problem

• Given a network of nodes and directed connections
  between them, is there a path through the network
  that begins with the start node and concludes
  with the end node visiting each node only once
  (Hamiltonian path")?
• Does a Hamiltonian path exist, or not?

49
Hamiltonian Path Problem (Cont.)
End city
Hamiltonian path does exist!
Detroit
Chicago
Boston
Start city
Atlanta
50
Hamiltonian Path Problem (Cont.)
Start city
Hamiltonian path does not exist!
Detroit
Chicago
Boston
End city
Atlanta
51
Hamilton Algorithm

• Hamilton Algorithm solves the Hamiltonian problem
• Given graph of n vertices or cities
• Produce or generate a set of randomly specified
  paths through the graph

52
Hamilton Algorithm (Cont.)

• For each path in the set
• Check whether that path starts at a start
  vertex
• and ends with the end vertex. If not,
  
• remove it from the set.
• Check if the path passes through
• exactly n vertices. If not, remove that
  path from
• the set.
• Keep only those that enter all cities once.

53
Hamilton Algorithm (Cont.)

• Finally, If the set is not empty, then report
  that there is a Hamiltonian path. If the set is
  empty, report that there is no Hamiltonian path.

54

• X D -gt B -gt A
• X B -gt C -gt D -gt B -gt A -gt B
• X A -gt B -gt C -gt B
• X C -gt D -gt B -gt A
• X A -gt B -gt A -gt D
• O A -gt B -gt C -gt D
• X A -gt B -gt A -gt B -gt C -gt D

55
Does a Hamiltonian path exist for the following
network?
56
Combinatorial Explosion

• The total number of paths grows exponentially as
  the network size increases
• (e.g.) 106 paths for N10 cities, 1012 paths
  (N20),
• 10100 paths!! (N 100)
• The Generation--Test algorithm takes forever.
  Some sort of smart algorithm must be devised
  none has been found so far (NP-hard).

57
Using DNA To Solve The Hamiltonian Path
58
The Founder

• In 1994, Leonard M. Adleman released an article
  in Science, in which he introduced the NP
  complete problem of the Hamilton path with DNA
  molecules.

59
Problem

• There are four cities, Bloomington, Indianapolis,
  Chicago, and St. Louis.
• A person driving a car from Bloomington to St.
  Louis trying to make a stop at each cities above
  exactly once. What would be the path for the
  person?

60
Solution
61
Encoding

• In this step, we basically just represent each
  city with a unique arbitrary DNA code.
• We also create a random Hamiltonian path from one
  city to another.

62
Graph

• The graph that representing the cities and the
  paths.

63
DNA Code Table

• The DNA and the Complement code for each city.

64
Paths Table

• The path from one city to another.

65
Generate All Possible Routes

• From the paths table, we will generate all the
  possible route from Bloomington to Chicago.
• Example
• Bloomington-gtIndy-gtSt.Louis-gtChicago
• TACGCCTATTAGGCTAAG

66
Sorting

• In this step we will sort the paths. We just pick
  the DNA path codes that have Bloomington as the
  initial city and Chicago as the final city by
  doing polymerase chain reaction.

67
Gel Electropheresis

• At this point, we already have a bunch of DNA
  strands that have the correct beginning and end,
  but we dont know anything about what is in the
  middle.
• Gel electropheresis works on the idea that DNA
  strands have electric charge. When we insert the
  DNA strands into it, it causes the DNA to move
  through the gel.

68
Gel Electropheresis (Continued)

• The shorter DNA strands move farther in the gel
  and the longer ones dont move as far.
• This means that the DNA strands spread according
  to the length.
• Then we will compare the a DNA strand that is
  known has the correct length to the DNA strands
  that are being tested to produce the DNA strand
  that has the correct length.

69
Magnetized Balls

• In this step, we need to test if the DNA strands
  that have the correct length visit each city and
  only once.
• This is done by attaching DNA strands to iron
  balls that match the various cities visited (the
  complements).
• First, we drop in the balls Bloomington attached
  and head up the mixture of DNA so that the
  strands break apart and attach themselves to the
  iron balls in the test tube.

70
Magnetized Balls (Continued)

• All the strands that do not visit Bloomington are
  not attached to the metal balls and can be
  siphoned off.
• The process is then repeated with iron balls
  representing all the cities that we are
  considering.
• After the last ball is tests the strands and the
  bad strands have been siphoned off, you are left
  with the answer.

71
Reporting the Correct Path

• The final step is to determine the sequence of
  the route through DNA sequencing and report all
  of the correct routes.
• There might be more that one answer. It depends
  on the number of cities involved and the number
  of ways you can visit the cities.

72
Architecture Kind Of
73
Dna Sequencing

• Several copies of input strand are made (100K-1M
  bp).
• Restriction enzymes are used to cut these strands
  into overlapping fragments.
• Flourescent tagging (the process of tagging
  things flourescently) and optical readout
  processes require a frag. Length of 500 bp.

74
Dna Sequencing(contd)

• The original sequence is reassembled by comparing
  the fragments (represented digitally with the
  atgc chars) and looking for a best fit overlap
• This process is repeated until original sequence
  is constructed.

75
Problems

• This is computationally expensive (involves
  sliding each strand past all the others to find a
  best fit).
• Another intense task is searching thru databases
  of genetic info with queries that are several
  thousand bp.
• The VLSI improves upon the speed for reassembly
  and sequencing by 3 Orders of Magnitude.

76
Who Does What?

• A readout mechanism stores fragment sequences
  into memory (??).
• The VLSI matches them.
• Then they are merged.

77
The Matching Alg
78
Isnt This Boring?
79
Architecture

• Shift Reg (S1) slides strand 1 past strand 2
  until their regs match (more later)
• Shifting prefetch buffer empties into S1 and
  avoids a data fetch for every cycle.
• Uses XNOR (matching exact) and XOR (matching
  complimentary) to compare
• Output is fed into a cons. AND treereturns true
  if all bp in both registers align.
• Bp index incremented when s1 is shifted.

80
Architecture

• Matching these small regions (contigs) are the
  core part of nearly every genetic processing
  algorithm.
• The hardware is capable of billions of bp/sec,
  which reduces the load on the software by several
  orders.

81
Speed Increase

• 2M register allows M simultaneous bp comparisons
• P aligns between the strands are checked simul.
  This means that the strands are shifted P bp
  every cycle.
• Pipelining allows MP bp comparisons to be made
  each clock cycle
• Memory is not a bottle neck

82
Thick Neck Memory

• 2 bit ATGC encoding allows 16 bp for 32 bit word
  fetch
• By using the shifting prefetch buffer, a fetch is
  only required every M/P cycles
• A 1M bp sequence, after fragmentation by 2
  restriction Enz. Is 2M bp.
• Encoded, this is only 500kB which will fit into
  L2 cache.

83
Dont Worry, This is the Last Slide

• Example 32 bit matching proc running at 500 MHz
  checked 4 aligns simul.
• 16 bp 4 aligns 500 106 32 109 bp
  comparisons/sec
• Memory requirement of one word every 4 cycles (16
  bp/4 aligns4), or a mem. Bandwidth requirement
  of
• 500MHz 4 bytes/4 cycles 500MB/s

84
Big NDs Conclusion
85
Pros

• DNA processors take much shorter time to perform
  computations too large to be run on electronic
  supercomputers.
• can solve more types of problems than electronic
  supercomputers.
• the potential for information storage in
  molecular computers follows the same trend as
  speed and efficiency DNA processors has an
  information storage density of 1 bit per cubic
  nanometer-a trillion times less space compare to
  1 bit per 1012 cubic nanometers information
  storage density with storage media of today, such
  as videotapes.

86
Pros (Continued)

• In energy efficiency, DNA processors can perform
  21019 power operations using one joule of energy
  compare to 1010 operations with supercomputer.
• In speed, DNA computers can perform 1000
  operations per second more than the fastest
  supercomputers(1012 operations per second) which
  means thousand million times slower than the DNA
  computers.

87
Pros (Continued)

• DNA computers use cheap, clean and readily
  available biomaterials rather than costly and
  often toxic materials that go into traditional
  microprocessors.

88
Cons

• DNA processors take longer time to multiply two
  100 digit integers(or other simple problems) than
  electronic supercomputers do.
• every operation with DNA computer, is somewhat
  random, that is, unlike the transistors in
  pentium, which reliably compute what they're
  supposed to

89
Cons (Continued)

• The components in the DNA computer are
  probablistic. for example, if the answer produced
  by pentium is 1, with DNA the answer is 1 90 of
  the time and 0 10 of the time.
• DNA computing is cumbersome because it is not
  entirely mechanized.
• It can be costly to make longer strands of DNA
  that encode more information, and programming DNA
  computers themselves can prove difficult because
  of the problems still inherent in manipulating
  DNA.

90
The Future

• Currently, molecular computing is a field with a
  great deal of potential, but few results of
  practical value. In the wake of Adleman's
  solution of the Hamiltonian path problem, there
  came a host of other articles on computation with
  DNA however, most of them were purely
  theoretical.

91
The Future (Continued)

• . Currently, a functional DNA "computer" of the
  type most people are familiar with lies many
  years in the future. But work continues in his
  article Speeding Up Computation via Molecular
  Biology ltftp//ftp.cs.princeton.edu/pub/people/rjl
  /bio.psgt Lipton shows how DNA can be used to
  construct a Turing machine, a universal computer
  capable of performing any calculation. While it
  currently exists only in theory, it's possible
  that in the years to come computers based on the
  work of Adleman, Lipton, and others will come to
  replace traditional silicon-based machines.