Artif Life Robotics (2008) 12:125128 DOI 10.

1007/s10015-007-0452-x

ISAROB 2008

ORIGINAL ARTICLE

Yoshiteru Ishida

Antibody-based computing: an application to the stable marriage problem

Received and accepted: July 23, 2007

Abstract Antibodies, among other things, are important components of the immune system. This paper proposes using the specic recognition capability exhibited by antibodies for computation, in particular, for solving the stable marriage problem, which has been studied as a combinatorial computational problem. Antibody-based computation is proposed by integrating the recognition capabilities of antibodies. The computation is carried out on an array form that is suitable not only for expressing stable marriage problems, but also for further integration to antibody microarrays. Key words Antibody-based computing Stable marriage problem Problem solving DNA-based computing Microarray

On the other hand, the post-genome era proved that the sequence information of the human genome alone is not sufcient, but a higher knowledge of their function must be revealed. The post-genome age naturally proceeds to study immune systems,7 focusing not only on components such as antibodies and MHC, but also on their systemic organization. After the demonstration of DNA-based computing by Adelman, protein-based computing was proposed by Hug and Schuler2 and extended by Balan and Krithivasan.8 This paper, in a similar spirit of synthetic biology and system biology, tries to construct an information processing device incorporating the specic recognition capabilities of antibodies. The stable marriage problem9,10 is used as a benchmark problem for demonstrating an array format implementation of antibody-based computation.

1 Introduction 2 Antibody-based computing: an introduction

After Adelman pioneered DNA-based computing in his seminal work, many researchers established that not only DNA, but also other macromolecules could have computational capability comparable to that of DNA.2 Elicits and Leibler3 and Sprinzak and Elowitz4 even demonstrated that a genetic circuit may not be a dream. Even synthetic multicellular systems have recently been studied,5 leading to synthetic biology. Possible drug production by engineering yeast6 has had a great impact on the area. These new bioengineering technologies have provided bioinformatics not only with new tools, but also with systemic views.
1

One of the major components propelling the information processing of the immune system is the specic recognition of antigens. The immune system is capable of recognizing even articially synthesized substances. Also, it can further classify substances into the self (those derived from the individual) and nonself. Among those bearing recognition capabilities, antibodies are undoubtedly important. Further, antibodies have been studied in great detail not only in theoretical biology, but also in bio-engineering. Antibodybased computing directly focuses on the recognition capability, and integrates it for problem-solving, including combinatorial problems. 2.1 Antibody-based computing and DNA-based computing Similar to DNA-based computing, antibody-based computing utilizes complementary matching between macromolecules: antibodies. Since the computational capabilities of DNA-based computing could be inherited by antibody-

Y. Ishida (*) Department of Knowledge-Based Information Engineering, and Intelligent Sensing System Research Center, Toyohashi University of Technology, Tempaku, Toyohashi 441-8580, Japan Tel. +81-53-244-6895; Fax +81-53-244-6873 e-mail: ishida@tutkie.tut.ac.jp This work was presented in part at the 12th International Symposium on Articial Life and Robotics, Oita, Japan, January 2527, 2007

126

based computing, we rather focus on the difference between them. Roughly, in the context of solving combinatorial problems, the difference may be claried by the correspondence between a complementary matching (as in the marriage theorem) and a stable pairing based on preference (as in the stable marriage problem). Afnity between antigens and antibodies can be measured and their intensities can be ordered (as formatted in an afnity matrix). That is, in contrast to Matching(DNA1, DNA2) = 1 (matched) 0 (not matched), Afnity(Antigen1, Antibody2) could vary from 0 (no agglutination) to 1 (highest agglutination). This difference would suggest that antibody-based computing could potentially implement an error tolerance that could not be implemented in DNAbased computing. The difference would further suggest that antibody-based computing may be extended to a problem solver if the adaptive mechanism of the immune system realized by antibodies and their maintenance is also involved.

Table 1. Landsteiners ABO blood group system Blood type A Antigen (agglutinogen) Antibody (agglutinin) A B B AB A, B None O None ,

Table 2. Afnity matrix. The circle indicates that the antibodyantigen reaction would occur if the antibodies in the column meet with antigens in the row Antibody Antigen A (anti-A) (anti-B) B

Table 3. Agglutination when the blood type in the column is transfused with the blood type in the row. A circle indicates that the blood type of the column would agglutinate when transfused to that of the column Blood type A A B AB O B AB O

3 Solving a combinatorial problem

3.1 Stable marriage problem In a naive form, the problem assumes n men and n women, with each member having preference lists for members of the opposite sex. A pair of a man Mi and a woman Wj is called a blocking pair if they are not a pair in the current solution, but Mi prefers Wj to his current partner, and Wj prefers Mi to her current partner as well. A matching between men and women with no such blocking pair is called stable. Having stated the stable marriage problem, it would be natural to think of the algorithm for antibody-based computing. That is, the stable marriage problem (SMP) may be mapped to an antigenantibody reaction, so that the preference order of each person in the SMP will be reected in the afnity intensity between an antibody and an antigen. After antibodies and antigens are so arranged, the solution of the SMP will emerge by observing the concentration of the agglutination. It should be remarked that the agglutination process could be any agglutination (not necessarily between antibodies and antigens) if their intensities are measurable and ordered. 3.2 Mapping a stable marriage problem to antibody-based computing As stated above, mapping a combinatorial problem to antibody-based computing can be done by composing antigenantibody compounds corresponding to a problem entity. Antibodies and antigens for a compound corresponding to a particular individual will be determined by considering her (his) preference list of men (women). Landsteiners ABO blood group system11 may be a familiar yet simple example. His blood type system is based on antigens (as agglutinogen) on red blood cells and anti-

Table 4. A trivial preference list for the two two stable marriage problem M1 W1 W2 1 2 M2 2 1 M1 M2 W1 1 2 W2 2 1

bodies (as agglutinin) in the blood serum. Table 1 shows the agglutinogen and agglutinin of each blood type. The afnity between antibody and antigen is shown in Table 2. Table 3 indicates the well-known incompatible transfusion among the blood types A, B, AB, and O. In this example, we map the relation that the woman Wi (the man Mi) prefers the man Mj (the woman Wj) to any other to the relation of whether the blood of Wi (Mi) would be agglutinate when the blood of Mj (Wj) was transfused. That is, if the woman Wi prefers the man Mj, most blood types should be so assigned that the type for Wi comprises antibody AbWi and antigen AgWi, and that for Mj of antibody AbMj and antigen AgMj, and the afnity Aff(AbWi, AgMj) is highest. For the trivial case when the preference lists of men and women are as in Table 4, a simple assignment would sufce: a man to type A and another man to type B for the woman who likes a man with type A to type B and for another woman type A (Fig. 1). It should be noted that the assignment to A for two men and to B for two women would not

127 Fig. 1. A blood-type assignment reecting the preferences

M1 M2

A B

W1

A W2

Fig. 2. A blood-type assignment reecting the preferences

M1 M2

O A

O W1 B W2

Table 5. A nontrivial preference list for the two two stable marriage problem M1 W1 W2 2 2 M2 1 1 M1 M2 W1 2 2 W2 1 1

work, since the assignment does not reect the preference of the men and women. In the nontrivial preference list shown in Table 5, one assignment would be type O to M1 and W1, type A to M2, and type B to W2 (Fig. 2). For other two preference lists (with the graph topologically different from those shown in Figs. 1 and 2), it is not possible to map the blood type with the above correspondence, and other compounds should be synthesized to realize the preference lists. These examples suggest a scheme for synthesizing antigenantibody compounds that realize mapping from given preference lists to the compounds. If the woman Wi prefers the man Mj to other men, the compound corresponding to Wi contains antibody AbWi and the compound corresponding to Mj contains antigen AgMj, which satises Aff(AbWi, AgMj) as the highest among other AgMj (j = 1 . . . n). If Mj is the second in the preference list of Wi, then Aff(AbWi, AgMj) must be second highest and so on. AgMj must realize the orders from women Wk other than Wi, hence the afnity Aff(AbWk, AgMj) must realize the order accordingly. (If AgMj alone cannot realize the order, then a new antigen realizing the order must be added to the corresponding compound. In fact, the above example also suggests that compounds can be composed of a set of antigens and a set of antibodies.) Constraints for selecting antibodies and antigens for a compound corresponding to a person can be summed up as follows: Aff(AbWi, AgMj) > Aff(AbWi, AgMk) if the woman Wi prefers Mj to Mk in her preference list for all Wi W, and for all distinct pairs Mj, Mk M; Aff(AbMi, AgWj) > Aff(AbMi, AgWk) if the man Mi prefers Wj to Wk in his preference list for all Mi M, and for all distinct pairs Wj, Wk W. 3.3 Solving a stable marriage problem with an array format An illustrative example with Landsteiners blood group suggests the following algorithm to solve the SMP with an

array format. In the array, row i and column j correspond to the compound for man i (i.e., AbMi and AgMi) and that for woman j (i.e., AbWi and AgWi). In other words, at the cross-point ij, two antigenantibody reactions between AbMi and AgWj (reecting the man is preference) and between AbWj and AgMi (reecting the woman js preference) will take place. Under the assumption that the concentration observed at each cross-point is proportional to both Aff(AbMi, AgWj) and Aff(AbWj, AgMi), the array can nd a stable matching by selecting one cross-point with the highest concentration from each row and column. This matching is certainly a stable one, for otherwise there must be a blocking pair Mk and Wl such that Aff(AbMk, AgWl) > Aff(AbMk, AgWp(Mk)) and Aff(AbWl, AgMk) > Aff(AbWl, AgMp(Wl)), where p(Mk) denotes a partner of Mk in the current matching. Then both concentrations at the cross-point kl are higher than those of kp(Mk) and those of p(Wl)l, reecting the afnities. Although obaining a stable matching shows some computational power, it can be solved in O(N2) time, where N is the number of men (and women). Gale and Shapley9 have invented a well-known algorithm for giving stable matches for man-oriented matching or woman-sided one. By further assuming that the concentration observed at a cross-point can reect the amount of antibodies imposed, the array is capable of obtaining any stable matching in the array from the man-oriented (man optimal and woman pessimal) matching to the woman-oriented (woman optimal and man pessimal) one. By equally increasing all the antibodies AbMi (i = 1 . . . n) (or equivalently antigens AgWj j = 1 . . . n) from a unit to , the matching would come close to the man-oriented one. Similarly, an increase of AbWi (i = 1 . . . n) will bias the matching towards the woman-oriented one. Since there are many variants of the stable marriage problem which are NP-hard, devising the array to solve these problems is challenging. Another challenge is to devise the array as a component of a problem solver that can deal not only with a particular problem, but also with similar problems, as the immune system has done.

4 Conclusion
We have shown that antibodies, a macromolecule of the immune system, with a specic recognition capability can be used for computation as a macromolecule DNA is used for DNA computing. Focusing on the stable marriage problem, and extending the ABO blood group system, it is shown that the antibody-based computation can be implemented on an antibody microarray.
Acknowledgments This work was supported in part by Grants-inAid for Scientic Research (B) 16300067, 2004. This work was also supported by the Global COE Program Frontiers of Intelligent Sensing, from the Ministry of Education, Culture, Sports, Science and Technology.

128

References
1. Adleman LM (2004) Molecular computation of solutions to combinatorial problems. Science 266:10211024 2. Hug H, Schuler R (2001) Strategies for the development of a peptide computer. Bioinformatics 17:364368 3. Elicits MB, Leiblerr SA (2000) Synthetic oscillatory network of transcriptional regulators. Nature 403:335338 4. Sprinzak D, Elowitz MB (2005) Reconstruction of genetic circuits. Nature 438:24 5. Basu S, Gerchman Y, Collins CH, et al. (2005) A synthetic multicellular system for programmed pattern formation. Nature 434: 11301134 6. Ro DK, Paradise EM, Ouellet M, et al. (2006) Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature 440:940943

7. Guo Z, Hood L, Malkki M, et al. (2006) Long-range multi-locus haplotype phasing of the MHC. Proceedings of the National Academy of Sciences, National Academy of Sciences, USA, pp 103118 8. Balan MS, Krithivasan K (2004) Parallel computation of simple arithmetic using peptideantibody interactions. Biosystems 76:303307 9. Gale D, Shapley L (1962) College admissions and the stability of marriage. Am Math Month 69:915 10. Irving RW, Leather P, Guseld D (1987) An efcient algorithm for the optimal stable marriage. J ACM 34:532543 11. Landsteiner K (1900) Zur Kenntnis der antifermentativen, lytischen und agglutinierenden Wirkungen des Blutserums und der Lymphe. Zentralbl Bakteriol 27:357362