FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Computing in Living Cells Ihab Sinno1 1
Nanoscale Science and Technology IMP Student
Abstract Generally speaking, a computer is an apparatus having some controllable input pattern that can be transformed by means of certain rules given for such a system. When properly interpreted, the transformation process can be understood as the execution of some desired computation, resulting in a detectable output pattern. As to current digital electronic computers, it is expected that the technology will saturate to its physical limits in the near-term future; hence, if other phenomena are not to be adopted, Moore’s law will shortly reach its inevitable end (performance-wise). Among the many diverse and unique proposals, computation in a living cell or organism sounds really interesting both technically and ethically. Current advancements in bioinformatics will surely help researchers to gain more control and wider scope over the whole subject. Moreover, living matter has a great advantage (or not) relying in the ability to self-manage, self-adapt and self-organize (so that real autonomous computations and communications may be carried-out to overcome the prospective complexity of problems). In this paper, some computational processes in living cells will be presented, with the main focus on the computation carried-out in ciliates.
1 / 19
Göteborg - Sverige
FKA 185
2 / 19
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Introduction When you think about computers, the image of the box of high-tech electronics comes instantly to your mind. However, in the nineteenth century the word computer meant a person who had been trained to perform routine calculations using little more than ink, quill pen, and paper. That seemed to be inefficient, since large-scale calculations needed a lot of accuracy and patience to be accomplished, while time was mostly exhausted by solving and checking different algorithms over and over. Therefore, many scientists tried to utilize various concepts and innovations in order to automate, accelerate and otherwise eliminate the hard work of extensive calculations. Successful inventions that lead to the development of today’s computers had their origin in the 1930's and early 1940's. In 1932, for instance, Vannevar Bush of the MIT completed a mechanical computer called the differential analyzer, which did calculus by rotating gears and shafts. Late in the 1930's Konrad Zuse of Germany, George R. Stibitz of the Bell Telephone Laboratories and Howard H. Aiken of Harvard University independently developed "electromechanical" computers, in which a series of relays represented numbers. The first electronic computer was designed between 1937 and 1942 by John V. Atanasoff and Clifford E. Berry and has come to be known as the Atanasoff-Berry Computer, or ABC. Atanasoff's digital computer mainly used capacitors and vacuum tubes in order to operate, and it easily attained an accuracy that was 1,000 times greater than it was possible with the differential analyzer, thus introducing the advantages of digital computers over analog systems.
The Atanasoff-Berry Computer
3 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Later on, more developed general purpose programmable digital computers were built, such as the Colossus (which was built in 1943 by the mathematicians Alan M. Turing and M. H. A. Newman), and the Electronic Numerical Integrator and Computer, or ENIAC (which was built in 1945 by John W. Mauchly and J. Presper Eckert). Thereafter, with the invention of the first transistor in 1948, computers started to be simpler, more compact, and much faster. In 1965 semiconductor pioneer Gordon Moore predicted that the number of transistors contained on a computer chip would double every year. This is now known as Moore’s Law, and it has proven to be somewhat accurate. The number of transistors and the computational speed of microprocessors currently double approximately every 18 months; components continue to shrink in size and have been becoming faster, cheaper, and more versatile. Today, the spread of computer technology into every aspect of modern civilizations ranks as one of the greatest achievements in human history. The ability to create, access, and share information and multimedia at the touch of a button has revolutionized the way we communicate, solve problems, plan, shop, and even play. However, with current technology limitations, how long will Moore’s law hold to satisfy our everlasting needs? In fact, on April 13, 2005, Gordon Moore himself stated in an interview that the law may not hold valid for too long, since transistors may reach the limits of miniaturization at atomic levels: “In terms of size [of transistor] you can see that we're approaching the size of atoms which is a fundamental barrier, but it'll be two or three generations before we get that far—but that's as far out as we've ever been able to see. We have another 10 to 20 years before we reach a fundamental limit. By then they'll be able to make bigger chips and have transistor budgets in the billions.” [1]
Intel’s 6nm transistor
Therefore, a true need is emerging for the utilization of different natural phenomena to develop new computational devices that are to enhance and may be replace current silicon technology.
4 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Turing Machines Turing machines are extremely basic symbol-manipulating devices which can be used to simulate the logic of any computer that could be possibly built. Those machines were described by Alan turning [2] in 1936, where according to the Church-Turning thesis, they mainly consist of: -
An endless tape which is divided into cells, where each cell contains a symbol from some finite alphabet
-
A movable head that can read/write the symbols on a given cell, one at a time
-
A state-register that stores the current state of the machine, given that possible states are finite
-
An embedded action table that tells the machine what symbol to write on the next cell, the position of the next cell as well as the next system-state (given the current state and the symbol it has just read)
A sketch for a Turing machine
Thus, throughout the quest to find suitable computation-capable phenomena, researchers have to inspect whether such phenomena can be used to build a universal turing machine in the first place. However, this condition by itself is never sufficient, since computational time, cost, space and energy are other crucial factors to be considered. Different chemical, physical, quantum physical and even biological phenomena are possible candidates for the development of computer components, and every year huge budgets are being invested to find suitable solutions. Such research is of a highly interdisciplinary type, a thing that is fortunately reflecting upon the improvement of different correlated fields. It should be noted that by the phrase ‘computation-capable’, it is not meant to have a device that can solve some quadratic equation, nor calculate the value of a certain function. However, any system that shows the ability to react in a certain intelligent non-random manner towards defined inputs can be considered as a computer component.
5 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Possible computer components
Computation in Biological Systems The boldest characteristic of biological systems and living matter in general is life itself. Regardless of the religious definition of life, living matter grows and evolves by reproduction (whether sexual or asexual), has a certain life-time and is doomed to death. During these different phases, the living system is always facing and evaluating different parameters, then referring to its embedded intelligence in order to take the corresponding decisions. Thus, it can be concluded that if all -or most of- the variables (inputs) for a given creature are well under control, it is possible to stimulate the system towards taking some decisions that can be readout (output); otherwise know as ‘computing’ using biological intelligence. The true power lying in such computation is that living systems process enormous amount of information on a daily basis, where most of it is done in a parallel manner. Moreover, just by understanding what kind of operations is being carried out in a certain natural process, one can construct laboratory made devices that simulate the same principle [3]. Other advantages for computing in living matter stem from its interdisciplinary nature; a thing that usually allows researchers to think out of the box (where plenty of options are still unveiled). This will also imply that different fields will be investigated simultaneously and using the same budget; hence, such research may reflect upon biology, medicine, chemistry, as well as information technology all in a single shot. The following figure for instance, illustrates the interdisciplinarity of BioMedical Informatics [Source: European Commission – Directorate General: Information Society].
6 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Another example that stresses out how important it is to think out of the box is the story behind DNA computing, (one sort of computation in living matter). Leonard Adleman was a mathematician and a computer scientist at the University of Southern California; however, in 1994 he wanted to model HIV, and he realized that he can’t talk the biologists’ conventional language. Therefore, he decided to learn some experimental techniques to be able to communicate with his colleagues. Upon reading about DNA machinery, he realized the striking similarities between DNA and Turing machines, and immediately started to investigate possible operations in his lab. Remarkably, on a very short notice (one week span) Adleman was able to solve one instance of Directed Hamiltonian Path (DHP) problem using DNA [4]. One more big advantage for computation in living cells is the possibility to develop autonomic computation and communication systems [5]. This is due to the fact that living matter and biological systems are considered to be very successful natural models of self-management and self-organization; both being crucial prerequisites for the design and control of next-generation communication networks (NGCNs). Those networks will be characterized by heterogeneity at different levels, connecting a large variety of users, media, processes and channels. Moreover, NGCNs will have to interact with its environment by collecting different parameters and taking appropriate decisions, either in a centralized or distributed fashion. These features will characterize a very complex computing and communication environment, where users will be highly mobile while relying on the end-to-end connection. Hence, the ability for such networks to self-adapt, self-organize and self-manage seems inevitable. Similarly, biological intelligence models have inspired researchers to attempt building an advanced robot that is capable of tracking a target, following the same control algorithms used by white blood cell Chemotaxis while tracking an intruder [6]. The investigated cells “neutrophils” are well known to be highly sensitive to low levels of chemical stimuli while being robust to noise. They are also capable of navigating unknown terrain; thus enjoying properties that are much desired in any autonomous clever robot.
7 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
A last example that I’ll be talking about extensively in the next pages is computation in “Ciliates”. Those unicellular organisms have shown remarkable capabilities for gene assembly using linkedlists and simple operations in a highly reliable process (accuracy is over 95%) [7]. Generally speaking, when researchers are approaching some biological system to investigate its computational capabilities, they are to follow these main steps: -
Comprehension of different related parameters, operations and aspects of the biological system
-
Building a molecular model
-
Building a mathematical model
-
Controlling the obtained model to yield certain results (sensitivity analysis)
-
Developing basic algorithms (easier with a computer model)
-
Testing the capabilities of their models by solving given problems
Ciliates The ciliates are one of the most important groups of kingdom protista; highly abundant in watercontaining places (lakes, ponds, oceans, soils, etc…). The name ciliate stems from the presence of hair-like protein organelles called cilia (Latin for eyelash), which are essential tools for the locomotion and feeding of the carnivorous unicellular organism. Ciliates are considered to be an ancient group of organisms, having existed over two an a half billion years ago. This group entitles about 10,500 genetically different organisms, each having its distinctive properties and sizes (some can be as large as 4mm). Despite being unicellular organisms, ciliates are rather considered to be one of the most complex protozoa in nature; being able to perform all main functions that larger organisms do (movement, sensitivity to the environment, water balance, food capture, reproduction, etc…).
Different kinds of Ciliates
8 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
As for the nuclear structure, Ciliates are considered to be special among eukaryotes, since they possess two types of nuclei; the micronucleus and the macronucleus (may have one or more of each nucleus). The micronucleus is mainly aimed-for as a germline nucleus; i.e. it is responsible for transferring genes to the daughter cells upon sexual reproduction. In it, DNA molecules are of very complex and long structures (millions of base pairs). On the other hand, the macronucleus is considered to be the somatic (bodily) nucleus in Ciliates. Therefore, it is in charge of organizing and managing the body of the cell itself. The genome found in the macronucleus is organized on short and simple plasmid molecules (200-20,000 base pairs), usually containing a single gene. This structural property leads to unprecedented optimization in the organism’s somatic processes (protein production at a very rapid pace) as well as for its adaptation capabilities (ciliates have virtually adapted to all types of water).
DNA molecules in the Micronucleus
DNA molecules in the Macronucleus
As to available reproduction methods, Ciliates can reproduce either sexually (conjugation) or asexually (fission or cloning).
In asexual reproduction, resulting daughter cells are genetically identical to the parent cell. For that, the micronucleus (MIC) undergoes mitosis while the macronucleus (MAC) simply pinches apart. As a result, the two daughter cells do not have exactly the same amount of DNA (average difference is 12% in Tetrahymena). However, that difference is not completely random; it is rather controlled via a special regulatory mechanism. Asexual Reproduction of Ciliates
9 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Sexual reproduction on the other hand is often suggested to be induced by the lack of food. In this process, two ciliates of complementary mating types come close together and develop a cytoplasmatic bridge (temporary union). Then, the micronuclei (MIC) of each organism will undergo meiosis while the macronuclei (MAC) start to disintegrate. In the meanwhile, gamete nuclei are exchanged through the union bridge so that a zygotic nucleus can be formed in each cell (by the fusion of the migratory and stationary gamete nuclei; thereby obtaining a new genetic information). Thereafter, the zygote micronuclei undergo one or more divisions yielding new macronuclei (MAC) and micronuclei (MIC) for the daughter cells. In other words, the old macronuclei are destroyed while the new ones (MAC) are produced from the zygotic micronuclei.
Sexual Reproduction of Ciliates
Importance of Ciliates in Computing After having this brief introduction to Ciliates, its time to know what’s special and important about them for the computational theory. However, in order to go through that discussion, there should be one last detailed description for the (MAC) and (MIC) genome structure.
10 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
The micronucleus is considered to be a germline nucleus (as mentioned before), where genes are encoded at irregular intervals along the chromosomes (giving a low density of coding in the DNA of each chromosome). The parts of the DNA molecule that carry the genetic data are referred to as MDSs (Macronuclear Destined Sequences), where as other nucleotide sequences are of no importance to the genetic data (used for disabling the gene and minimizing possible mutation probability). These redundant parts are known as IESs (Internally Eliminated Sequences, usually less than 100 nucleotides in length). In other words, genes inside the micronucleus are broken and scrambled into pieces of MDSs separated by the non-coding ‘useless’ IESs where typically, a ciliate’s genome can contain more than 100,000 IESs. Conversely, the macronucleus carries DNA as short plasmids, most of which are taken by a single gene (the shortest known DNA molecules in nature). Those genes are composed of the micronuclear MDSs assembled in the orthodox order.
Now one can start making sense of the importance of ciliates as computational tools. It all relies in the ability of ciliates to transform long, encoded and scrambled DNA molecules (millions of bps) embedded in the MIC into short ordered plasmids (200~20,000 bps) carrying correct genes in the MAC during sexual reproduction (with a very high accuracy over 95%). Hence, ciliates must be using some elegant detecting and matching schemes, along with tremendously complex pointers and linked list as data structures.
11 / 19
Göteborg - Sverige
CHALMERS UNIVERSITY OF TECHNOLOGY
FKA 185
MAY 2006
Modeling As for now, it’s been mentioned that in the MIC-MAC transformation, all IESs are precisely excised from the genes. Simultaneously, all MDSs are spliced together to form transcriptionally competent genes. However, it’s time to know how to model that computational tool in order to comprehend some basic natural examples. -
Each MDS (except the first and the last) has the following structure: Mi = (πi , μi , πi+1) ; where πi is the incoming pointer of Mi , μi is the body, and πi+1 is the outgoing pointer
-
M1 = (b , μ1 , π2)
-
Mk = (πk , μk , e)
-
The outgoing pointer of Mi coincides with the ingoing pointer of Mi+1 (same nucleotide sequence)
-
The MDSs will be spliced together on their pointer to form the MAC genes
-
Each MDS can be represented by its pair of pointers
-
The same nucleotide sequence that forms a pointer can occur many times along the DNA molecule; however, its only a pointer when it is placed at the border of an MDS or IES
A section of the MIC gene encoding βTP in Sterkiella histriomuscorum
Then, as for gene assembly, the goals can be simply summarized by: -
Remove the IESs
-
Unscramble (place in the orthodox order) the MDSs
-
Ligate the MDSs
-
Excise the gene from the chromosome
-
Add telomeres
-
Multiply the DNA molecule to reach the required number of copies (depending on the species)
12 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Regarding the basic molecular operations used for gene assembly, there exist three different kinds of operations: (1) Ld (excision) - Also known as “loop, direct repeat, excision” process - The operation is applied to DNA sequences having a direct repeat pattern of a pointer π - First, the molecule folds as a loop to align the two occurrences of a pointer π - Second, enzymes will cut on the pointer sites - Third, hybridization takes place - Type of the fold: loop - Type of repeat: direct repeat - Type of operation: excision
Ld operation, it excises unnecessary DNA parts
(2) Hi (excision/reinsertion) - Also known as “hairpin, inverted repeat, excision/reinsertion” process - The operation is applied to those DNA sequences having an inverted repeat pattern of a pointer π - First, the molecule folds on itself as a hairpin to align the corresponding pointers π - Second, enzymes will cut on the pointer sites - Third, hybridization takes place - Type of the fold: hairpin - Type of repeat: inverted repeat - Type of operation: excision followed by reinsertion
13 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Hi operation, it inverts a DNA sequence
(3) Dlad (excision/reinsertion) - Also known as “double-loop, alternating direct repeat, excision/reinsertion” process - The operation is applied to those DNA sequences having an alternating direct repeat pattern of two pointers π1 and π2 - First, the molecule folds as a double loop to align the two occurrences of π1 and the two occurrences of π2 - Second, enzymes will cut on the pointer sites - Third, hybridization takes place - Type of the fold: double loop - Type of repeat: alternating direct repeat - Type of operation: excision followed by reinsertion
Dlad operation, it exchanges the places of two different DNA sequences
Other issues: -
In a successful assembly, no MDS is lost
-
In the macronucleus, all molecules are linear; even if the gene is assembled circularly, enzymes will cut it out and add telomeres to make it linear
-
Molecular mechanisms that execute DNA processing must perform two major functions: an accurate recognition for the cutting sites, and supplying suitable enzymes to cut and splice the DNA
14 / 19
Göteborg - Sverige
FKA 185
-
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Recognition can be traced back to two explanations: the detection of differences in the TA content of neighboring regions, or the old MAC acts as a template while pointers are just used for splicing
Examples To have a feeling for the complexity at which such operations are conducted, it should be mentioned that huge amounts of MDSs may have to be ordered, spliced, and then excised with precision. A real life example is the O.Trifallax Ciliate, where it contains in its genome more than 100,000 IESs that have to be excised from more than 100,000 scrambled MDSs (that are to be ordered); amazingly, a survival rate of more than 98% is achieved upon mating. The following figures will illustrate a simple example of gene assembly carried out for actin I on S. Nova:
15 / 19
Göteborg - Sverige
FKA 185
16 / 19
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
On the other hand here is a more realistic parallel assembling representation for actin I on S. Nova:
17 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
Conclusion In conclusion, it was shown throughout the previous text how it is possible to benefit from different biological and living systems for improving computational processes. As for the investigated example of “computing in Ciliates”, several shortcomings can be predicted: -
the contexts that are needed to simulate a Turing universal machine depend essentially on the machine it self; hence, they may be far away from the MDS-IES junctions that the ciliates may recognize
-
Many copies of the original MIC string (chromosome) are needed to perform a specific program. However, it is still difficult to imagine how one can create any number of DNA molecules inside the micronucleus of a ciliate
-
Computational approach needed may be different and simpler for “basic” non universal computations
Nevertheless, researchers will continue to investigate biological intelligence at different levels in an attempt to find out how nature really works it. Such a fact reflects how ambitious, curious and controlling a human being is; in the end, nature defines our variables, and a decision has to be made…
18 / 19
Göteborg - Sverige
FKA 185
CHALMERS UNIVERSITY OF TECHNOLOGY
MAY 2006
References [1] Manek Dubash, Techworld (April 13 2005) [2] Rolf Herken, “The Universal Turing Machine – A Half-Century Survey”, Springer- Verlag [3] Zoran Konkoli, “FKA185, Nano-scale reactors and networks”, CTH (2006) [4] Zoran Konkoli, “FKA185, DNA Computing”, CTH (2006) [5] I. Carreras, I. Chlamtac, F.D. Pellegrini, C. Kiraly, D. Miorandi, and H. Woesner, “A biological approach to autonomic communication systems”, Springer-Verlag (2006) [6] M.D. Onsurn and A.P. Arkin, “autonomous mobile robot control based on white blood cell chemotaxis”, Springer-Verlag (2005) [7] “Computational processes in living cells” spring course oferred by Ion Petre at Åbo Akadimi, Finland (2005)
Active groups mentioned throughout the course given in reference [7] are: -
Intramolecular model for Ciliates: Ion Petre (Abo Akadimi-Finland), Tero Harju (Abo Akadimi-Finland), David Prescott (Colorado University-USA), Andrzej Ehrenfeucht (Colorado University-USA) and Grzegorz Rozenberg (Univ. of Leiden – Hollend)
-
Intermolecular Model for Ciliates: Laura Landweber (Princeton-USA) and Lila Cari (University of Ontario, Canada)
19 / 19
Göteborg - Sverige
