1. DNA and Quantum Computers Trends in Computational Science DNA and Quantum Computers Lecture 2 (cont)

2. Sources 1. Nicholas Carter 2. Andrea Mantler, University of North Carolina 3. Michael M. Crow, Executive Vice Provost, Columbia University. 4. Russell Deaton, Computer Science and Engineering, University of Arkansas 5. Julian Miller 6. Petra Farm, John Hayes, Steven Levitan, Anas Al-Rabadi, Marek Perkowski, Mikael Kerttu, Andrei Khlopotine, Misha Pivtoraiko, Svetlana Yanushkevich, C. N. Sze, Pawel Kerntopf, Elena Dubrova

3. Dominating role of biology and system science

4. Outline Trends in Science Programmability/Evolvability Trade-Off DNA Computing Quantum Computing DNA and Quantum Computers

5. Introduction What’s beyond today’s computers based on solid state electronics? Biomolecular Computers (DNA, RNA, Proteins) Quantum Computers Might DNA and Quantum computers be combined? Role of evolutionary methods.

6. Scientific questions are growing more complex and interconnected. We know that the greatest excitement in research often occurs at the borders of disciplines, where they interface with each other.

7. Computers and Information Technology No field of research will be left untouched by the current explosion of information--and of information technologies. Science used to be composed of two endeavors--theory and experiment. Now it has a third component: computer simulation, which links the other two. - Rita Colwell

8. Government Initiatives on Information Technology Interdisciplinary teams to exploit advances in computing –Involves computer science, mathematics, physics, psychology, social sciences, education Focus on: –Role of entirely new concepts, mostly from biology –New technologies are for linking computing with real world - nano-robots, robots, intelligent homes, communication. –Developing architecture to scale up information infrastructure –Incorporating different representations of information (visual, audio, text) –Research on social, economic and cultural factors affected by and affecting IT usage –Ethical issues.

9. The Price of Programmability Michael Conrad Programmability: Instructions can be exactly and effectively communicated Efficiency: Interactions in system that contribute to computation Adaptability: Ability to function in changing and unknown environments

10. What is nanotechnology? –Any technology below nano-meter scale Carbon nano-tubes Molecular computing Quantum computing Are we going there? –Yes, a technology compatible with existing silicon process would be the best candidate. Is it too early for architecture and CAD? –No CAD problems in nanotechnology

11. Nanotechnology Atoms <1 nm DNA ~2.5 nm Cells thousands of nm We are at the point of connecting machines to individual cells

12. Federal Initiative: Nanotechnology Interdisciplinary ability to systematically control and manipulate matter at very small scales –Involves biology, math, physics, chemistry, materials, engineering, information technology Focus on: –Biosystems, structures of quantum control, device and system architecture, environmental processes, modeling and simulation

13. Biocomplexity Planet Biome Ecosystem Community Habitat Population Organism Organ Tissue Cell Organelle Molecular Atomic REDUCTIONISM INTEGRATION

14. Multidisciplinary

15. Human Genome Sequence Next race: annotation –Pinpoint genes –Translate genes into proteins –Assign functions to proteins DNA chipGenomic tool example: DNA chip –Array of genetic building blocks acts as “bait” to find matching DNA sequences from human samples Entire yeast genome on a chip

16. DNA Computers Massive ParallelismMassive Parallelism through simultaneous biochemical reactions Huge information storage density In Vitro Selection and Evolution SatisfiabilitySatisfiability and Hamiltonian Path

17. DNA Computing (Adleman, 1994) DNA is the hereditary molecule in every biological cell. Its shape is like a twisted rubber ladder (i.e. a double helix). The rungs of the ladder consist of two bonded molecules called G, C, A, T. bases, of 4 possible types, labeled G, C, A, T. G can only bond (pair off) with C, and A with T. What is DNA? Basic Coding A = adenine T = thymine C = cytosine G = guanine

18. Double Helix

19. Base Pairing

20. strand (string) If a single strand (string) of DNA is placed in a solution with isolated bases of A, G, C, T, then those bases will pair off with the bases in the string, and form a complementary string, e.g. G A T T C A G A G A T T A T C T A A G T C T C T A Strands and Pairing Off DNA Code Sticky Ends

21. DNA operations 1 Separating DNA strands (denaturation) Binding together DNA strands –(renaturation or annealing) Completing sticky ends Synthesizing DNA molecules

22. DNA operations 2 ShorteningShortening DNA molecules CuttingCutting DNA molecules

23. DNA operations 2 LinkingLinking DNA molecules InsertingdeletingInserting or deleting short subsequences MultiplyingMultiplying DNA molecules

24. DNA operations 3 Filtering Separation by length Reading

25. The Hamiltonian path problem The Hamiltonian path problem: In a directed graph, find a path from one node that visits (following allowed routes) each node exactly once. This complementary bonding can be used to perform computation, e.g. a version of the traveling salesperson problem (TSP), called the Hamiltonian Path problem

26. Hamiltonian path as an example of graph theory problem This kind of problems are abstracted as graphs. Graphs has nodes and edges. Graphs are oriented (like the above) and non-oriented.

27. Start with a directed graph G (i.e. the edges between nodes are arrows) at node A, and end at node B. The graph G has a Hamiltonian Path from A to B if one enters every other node exactly once. E.G. for the directed graph G 1 shown, a b c A B de A solution, (G 1 ’s Hamiltonian Path) is A => c => d => b => a => e => B G1G1 Example of a solution to HP problem :-

28. Adelman’s DNA algorithm for Hamiltonian Path Input: n start node A end node B. –A directed graph with n nodes including a start node A and an end node B. Step 1. Generate graphs of the above form, randomly, and in large quantities (Generate random paths through the graph). Step 2. Remove all paths that do not begin with start node A and end with end node B. Step 3. If the graph has n nodes, then keep only those paths that enter exactly n nodes. Step 4. Remove any paths that repeat nodes Step 5. If any path remains then answer “yes”otherwise answer “no”. This is a nondeterministic algorithm.

29. DNA can implement this algorithm! (Uses 10 15 DNA strings) random 20 base string Step 1 : To each node “i” of the graph is associated a random 20 base string (of the 4 bases A,G,C,T), e.g. TATCGGATCGGTATATCCGA Call this string “S-i”. (It is used to “glue” 2 other strings, like LEGO bricks). 1 glue 2 DNA Computer for this problem Representation of nodes

30. For each directed (arrowed) edge (node “i” to node “j”) of the graph, associate a 20 base DNA string, called “S-i-j” whose - a) left half is the DNA complement (i.e. c) of the right half of S-i, b) right half is the DNA complement of the left half of S-j. Step 2 : The product of Step 1 was amplified by “Polymerase Chain PCR Reaction” ( PCR ) using primers O-A and (complement) cO-B. Thus, only those molecules encoding paths that began with node A and ended with node B were amplified. S-i-j Glue S-jGlue S-i 20 bases 10 bases Representation of edges

31. Litigation

32. Implementing the algorithm with DNA Create a unique sequence of 20 nucleotides to represent a node. Similarly create 20 nucleotide sequence to represent the links between nodes in the following way:

33. Step 1. Generate random paths

34. Step 2a. Denature and add node 0 primer and node 6 anti-primerStep 2a. Denature and add node 0 primer and node 6 anti-primer Recall: denature = separate strands

35. Step2b: PCR amplifies 0-6 strands

36. Step3: Find paths with 7 nodes The product of step 2 was separated according to length by electrophoresis. The DNA with 140 base pairs was extracted, PCR amplified, subjected to electrophoresis a few times to purify sample PCR is a technique in molecular biology that makes zillions of copies of a given DNA (starter) string. Step 3 : The product of Step 2 was run on an gel, and the 140-base pair (bp) band (corresponding to double-stranded DNA encoding paths entering exactly seven nodes) was extracted.

37. Step4: extract paths that contain all nodes The product of step 3 was denatured. Magnetic beads with complementary node sequences (nodes 1 to 5) were obtained. The product was successively filtered by annealing with solutions containing single complement node beads Step 4 : Generate single-stranded DNA from the double-stranded DNA product of Step 3 and incubate the single-stranded DNA with cO-i stuck to magnetic beads. Only those single-stranded DNA molecules that contained the sequence cO-a (and hence encoded paths that entered node a at least once) annealed to the bound cO-a and were retained

38. Step5: PCR amplify remaining product See if any product left. Actually the exact node sequence in the path can be obtained by a process called graduated PCR This process was repeated successively with cO-b, cO-c, cO-d, and cO-e. Step 5 : The product of Step 4 was amplified by PCR and run on a gel (to see if there was a solution found at all).

39. Problems with Adelman’s Appraoch to DNA computing Solving a Hamiltonian graph problem with 200 nodes would require a weight of DNA larger than the earth! What algorithms can be profitably implemented using DNA? What are the practical algorithms? Can errors be controlled adequately?

40. . DNA Computers

41. This work took Adleman (the inventor of DNA computing, 1994) a week. See November 11, 1994 Science, (Vol. 266, page 1021) As the number of nodes increases, the quantity of DNA needed rises exponentially, so the DNA approach does not scale well. The problem is NP-complete. But for N nodes, where N is not too large, the 10 15 DNA molecules offer the advantages of massive parallelism. Summary on Adleman

42. DNA Computers

43. Problems with DNA Computers

44. Can be adaptable through enzymatic action Hard to program because of hybridization errors Not very efficient because of space complexity

45. DNA Self-Assembly

46. Molecular Computing Building electronic circuits from the bottom up, beginning at the molecular level Molecular computers will be the size of a tear drop with the power of today's fastest supercomputers Single monolayer of organically functionalized silver quantum dots Journal of Physical Chemistry, May 6, 1999

47. Molecular Computing as an Emerging Field Interdisciplinary field of quantum information science addresses atomic system (vs. classical system) efficiency and ability to handle complexity –Involves physics, chemistry, mathematics, computer science and engineering Quantum information can be exploited to perform tasks that would be nearly impossible in a classical world Observing quantum interference

48. Quantum Computers Different operating principles than either DNA or conventional computers Coherent superposition of states produces massive parallelism Explores all possible solutions simultaneously Prime Factorization, Searching Unsorted List

49. Quantum Computers Qubits: |Q> = A |0> + B |1> |A| 2 + |B| 2 = 1 P(0) = |A| 2, P(1) = |B| 2 1 0

50. Quantum Computers: CNOT Gate

51. Quantum Computers Very efficient because of superposition of exponential number of states Can be programmed Not adaptable Must be isolated from the environment

52. DNA Assembly of Quantum Circuits

53. DNA NMR Computers http://rabi.cchem.berkeley.edu/~kubinec/slideshow1/slideshow/sld013.htm

54. DNA Qubit

55. Research Issues What is the state-of-art? –A lot of funding available! –Lots of experimental research on device level Molecular RAMs Carbon-film memories –Limited activity of higher levels of design –Lack of communication between physicists/chemists and architecture/CAD engineers

56. What could the CAD community contribute at this stage? Identifying which properties we need to build circuits –Composability/cascadability: (x')' = x –Gain for signal restoration Restoring logic (molecular amplifiers) Error-correcting techniques in quantum computing Techniques for building reliable circuits from unreliable components –What logic abstractions do we need for that? –How much can be borrowed from the existing fault-tolerant design techniques?

57. Conclusions DNA, Quantum Computers and other nano- technologies have great potential. Serious technical barriers need to be overcome. These technologies have complementary properties. Work together to have programmability, efficiency, and adaptability It is not too early to think about CAD, architectures, and algorithms. Past (thousands), present (50 years) and future (few years) technologies of computing. Be braveBe brave, have a perspective.

58. Reading Assignment 1. Read slides to lectures 1, 2 and 3 from my WebPage. 2. Read Chapter three (Introduction to Computer Science) from Nielsen and Chuang. 3. Read Chapter 1. Sections 1.1, 1.2, 1.3, 1.4.1, and 1.4.2. You may expect a very short quizz next week. This is the end of Lecture 2.