Doctor of Philosophy
Research Proposal

Shih Ching Fu
School of Computer Science & Software Engineering
The University of Western Australia
35 Stirling Highway, Crawley, W.A. 6009, Australia.
scfu(at)csse.uwa.edu.au

August 2003

A.  Proposed Study

1.  Provide a title.

Evolution Strategies for ab initio Protein Structure Prediction

2.  Higher degree regulation 64 specifies that a PhD study must make a ``substantial and original contribution to scholarship, for example through the discovery of new knowledge, the formulation of theories or the innovative re-interpretation of known data and established ideas''. In what way is the proposed study expected to fulfill this requirement?

Evolutionary Computation

Evolutionary computation is the branch of computer science that deals with solving real-world problems by taking inspiration from evolutionary theory. It has been demonstrated that capturing the natural processes of inheritance, mutation, competition, and selection into algorithmic form will result in a robust method of optimization [15]. These algorithms work on the principle of using a population of candidate solutions that explores a search space, rather than relying upon a single candidate. Starting with an initial pool of candidates, evolutionary processes such as mutation and fitness proportionate selection are used to produce new generations of individuals, which under the influence of a fitness function, become more suited to solving the problem in question.

There are several sub-branches of evolutionary computation, including genetic algorithms, genetic programming, evolution strategies, and evolutionary programming. These branches differ primarily in their representation of candidate solutions and the operators used to derive solutions. The manipulation of candidate solutions by operators such as mutation and recombination heavily influences how effectively the algorithm explores the search space [46].

Holland [24] first introduced genetic algorithms in 1975. Initially proposed as an adaptive search technique, genetic algorithms have been frequently used in optimization problems. In these algorithms, strong emphasis is placed on the genetic encoding of potential solutions and the use of genetic operators. Typically, candidate solutions are represented by bit strings or chromosomes. The mapping of candidate solutions into these chromosomes emulates the mapping between phenotypes and genotypes as found in nature. The operators used in genetic algorithms reflect those found in natural reproduction, namely mutation and crossover.

Genetic programming is an extension of genetic algorithms. Koza [28] first used the term genetic programming to describe the use of genetic algorithms over syntactically correct computer programs represented as tree-structured chromosomes. Genetic programming can be likened to genetically generating programs and optimizing these hierarchical programs for specific problem domains. Such programming has been used for developing artificially intelligent agents [16].

The concept of an evolution strategy was initially proposed by Rechenberg and Schwefel [2]. Evolution strategies differ from genetic algorithms in that they do not aim to imitate the biology of natural evolution. Rather, the representation of individuals is closer to the natural representation specified by the problem. The main operators used in evolution strategies are mutation and recombination. Candidates are often represented as tuples of real-values and mutations are usually introduced as Gaussian perturbations. Recombination is analogous to the genetic crossover operations found in genetic algorithms. Evolution strategies have been successfully applied to many engineering applications such as heat exchange network design [21], ore crusher design [3], and jet nozzle optimization [27].

Evolutionary programming looks at evolving finite state machines. Fogel [17,16] proposed using the processes present in natural evolution to design intelligent agents, these agents taking the form of computer programs, which in turn were represented as finite state automata. These agents could then be used for prediction, control, or perhaps classification tasks - a form of artificial intelligence.

Given the similarities in their underlying framework, these branches of evolutionary computation are collectively referred to as evolutionary algorithms (EAs). There is no restriction that the operators or representations in one branch of evolutionary computation cannot be applied to another branch. Often the boundaries different evolutionary techniques are historical and easily blurred as described by De Jong and others [25,2,15].

A growing field within evolutionary algorithms deals with multi-objective optimization [18]. It is common that the interplay between several objectives of a problem means that there does not exist a single optimum solution. This is because the optimization of one objective is typically at the expense of another. It is then desirable to find the whole collection of solutions representing the best compromises between the objectives. There are many applications of single-objective evolutionary algorithms which can more naturally be recast into multi-objective evolutionary algorithms [47,19,9].

Protein Structure Prediction

The basic building blocks of life are proteins [5]. Almost all bodily functions of all organisms on earth involve one protein or another. However, despite the vast RNA databases gleaned from genome mapping projects providing us with the amino acid sequence of thousands of proteins, there is no simple way of predicting the function of these proteins. The Protein Folding Problem, a `Grand Challenge Problem' of science involves the prediction of a protein's function in the body given only its amino acid sequence [36].

It is believed that the function of a protein is heavily dependent on its three-dimensional native conformation [41]. Therefore, the Protein Folding Problem can be recast as the prediction of a protein's three dimensional or tertiary structure. Note that it is possible to determine the tertiary structure of existing proteins using methods such as X-ray crystallography and nuclear magnetic resonance imaging (NMR), but these techniques are time consuming, error prone, and expensive. It is therefore desirable to be able to generate protein structures via simulation.

There are two main approaches to protein structure prediction (PSP): knowledge-based and ab initio. The first approach focuses on using an existing database of known protein structures and through amino acid sequence comparisons, the structure of the protein may be extrapolated. Obvious shortcomings of this approach are the poor predictive results for proteins comprising new or unknown folds not existent in the database. This approach also provides no insight to the process of protein folding as it occurs in nature. The second approach, ab initio structure prediction, has a more ambitious aim of predicting three-dimensional conformations given only amino acid sequences. Currently, the knowledge based approach to PSP has been more successful but it lacks the generality that ab initio techniques would have.

Contribution to scholarship

The focus of my research will be examining existing technologies in evolutionary computation and applying them in novel ways to protein folding. Possible avenues include secondary protein structure prediction, tertiary protein structure prediction, amino acid sequence alignment, and primary (amino acid) structure matching. Most of the current simulators used in de novo protein design have examined a reduced problem with simplified side chain representations or simplified molecular interaction models, usually to limit computation time. The use of an EA will perhaps decrease computation times and improve prediction accuracy by increasing the coverage of the search space.

At a higher level, evolutionary technologies such as multi-objective optimization, may find application in protein folding. In such a case, a formal methodology for implementing a multi-objective EA may be realized and applied to other application domains. Other technologies and techniques include dealing with noise in candidate fitness evaluations, parallelization of the algorithm implementation, or extension of the self-adaptation paradigm where candidate solutions may hold more than just their mutation parameters, such as the simulation of chaperon proteins. It is possible that not only will the application of evolutionary techniques better help us understand the protein folding process but also the action of natural evolution itself.

B.  Research Plan

1.  The research topic should have been defined to the mutual satisfaction of the candidate and the supervisor(s). The supervisor(s) should assist in preparing a framework for the research with the time estimates for the completion of its various phases bearing in mind that the maximum period of candidacy is four years (full-time). This will ensure that all parties have a template for monitoring the progress of the research and a positive orientation to the timely completion of the thesis.

  1. Literature Review
    A literature review of the field of evolutionary computation has been completed. The purpose of this review is to identify new areas of research within evolutionary computation, such as multi-objective optimization, as well as identify new application domains where EAs can be applied in a novel manner. A field of particular interest is protein structure prediction.

  2. Investigation into automated map labelling
    As an introduction into the area of multi-objective evolutionary algorithms, the problem of automated map or graph labelling will be examined. Currently this problem has been tackled with single objective EAs. An extension of this work into the realm of multi-objective EAs will hopefully provide a generalized framework upon which to build other multi-objective optimization applications.

  3. Detailed research into protein structure prediction
    A detailed review of the state of the art in the protein structure prediction domain will need to be carried out. Such a review will require external collaboration from outside the School of Computer Science & Software Engineering, perhaps from Biochemistry or Biophysics. The problem areas of protein structure prediction that are conducive for search strategies will need to be identified in preparation for the application of EAs. From readings of the literature thus far, it appears that evolutionary techniques could be used for genetic sequence matching, as well as secondary and tertiary protein structure prediction.

  4. Design of framework for an evolutionary algorithm and the protein structure prediction problem
    Design considerations for an evolutionary algorithm for the protein folding problem will include:
    • Deciding on a representation for unfolded proteins.

    • Discovery of a molecular interaction model with which candidate fitnesses will be measured.

    • Design of novel mutation and recombination operators for effective exploration of the conformation search space.

    • Design of candidate selection policies.

    • Deciding how the assessment of native conformation predictions will be done, that is, defining the algorithm termination criteria.

    Particular notice will be given to the ideas behind the LINUS hill climbing algorithm [39] and how it can be recast into a population based algorithm.

  5. Implementation of protein structure prediction EA
    Implementation and testing of the above mentioned algorithm framework is estimated to be completed in early 2005.

  6. Experimentation and assessment of EA effectiveness
    Experiments will need to be conducted to assess any improvements that may result from using evolutionary techniques over existing algorithms as well as provide an insight to where shortcomings lie. Further experiments will be conducted to investigate the impact of design decisions, such as those mentioned above, on the results of the EA.

  7. Thesis composition
    The writing of my thesis is expected to take approximately six months. Submission is anticipated to be during the end of 2005.

2.  The specific aims of the project - the problem(s) it hopes to solve or particular question(s) it will answer.

  • Identify possible computational problem areas of the Protein Folding Problem.

  • Apply search based techniques like single or multi-objective evolutionary algorithms to these problem areas.

  • Assess the usefulness of evolutionary algorithms compared to other algorithm paradigms in the Protein Folding Problem.

  • Generalize the techniques found from protein folding experiments to other similar graph based problems such as secondary protein structure prediction, sequence matching, and other drug manufacture issues.

3.  The methods to be used or the approach to be taken. What similar projects have been undertaken here or elsewhere; have similar methods been used before?

Evolutionary algorithms have already been applied to protein structure prediction; most commonly tertiary structure prediction using genetic algorithms. Here, the protein structure prediction problem is recast as an optimization problem where the overall conformational energy of a protein molecule is minimized. It was postulated by Anfinsen that the final conformation of a protein has, to a first approximation, the global minimum molecular energy [1]. Consequently, most ab initio prediction methods use this fact to search through the millions of possible folded protein structures to find the one with lowest energy; that is, search through the space of all possible folds to find the molecule with minimum energy. However, recent theories suggest that the final energy of a conformation is not the determining factor in protein folding. Instead, the intermediate energy states of subsequent folds determine which particular occur [39].

The simulations done by Srinivasan and Rose [40] assume that it is not the final conformational energy of a protein that is minimized, rather it is the intermediate partially folded stages of the protein that desire minimum energy. Although they have used a hill climber algorithm called LINUS to generate predictions whose conformations do not necessarily have global minimum energy, their results are promising. The LINUS algorithm is a purely ab initio method of protein structure prediction that uses partially folded energy states to determine the most desirable folds with no input apart from the primary amino acid structure.

The LINUS algorithm however, is very time intensive and does not outperform knowledge based approaches [33]. In light of this, the performance of the folding simulations may be improved by applying evolutionary algorithms to the search based components of the simulation. To perhaps improve the accuracy of predictions, a multi-objective EA could be used where it has been suggested that not only is molecular energy an objective, but so is entropy, symmetry, and hydrophobic/hydrophilic interaction. Further performance gains may be obtained by parallelizing the search procedure given that candidate fitness evaluations can be carried out independently [15]. This parallelization may be necessary considering the size of the conformation search space.

4.  What efforts have been made to ensure that the project does not duplicate work already done?

I have conducted searches of online and hardcopy literature and are not aware of any similar research. This research relies upon extending research in an original fashion.

Ongoing review of contemporary literature and consultation with the evolutionary algorithm research community (such as conference attendance) will be essential to keeping my research original and up to date.

C.  Scholars

1.  Identify some leading scholars in the field, particularly some whose published work you have had occasion to study. If possible, include at least one from Australia.

David B. Fogel
(dfogel@natural-selection.com)
Natural Selection Inc.
3333 North Torrey Pines Ct., Suite 200, La Jolla, CA 92037 USA
Hans-Paul Schwefel
(hps@udo.edu)
Department of Computer Science, University of Dortmund
D-44221, Dortmund, Germany
Hussein A. Abbass
(abbass@cs.adfa.edu.au)
School of Computer Science, University College, Australian Defence Force Academy, University of New South Wales
Canberra, ACT 2600
Natalio Krasnogor
(natalio.krasnogor@nottingham.ac.uk)
School of Computer Sciences and IT
University of Nottingham, Nottingham, NG8 1BB, United Kingdom
George D. Rose
(rose@grserv.med.jhmi.edu)
Department of Biophysics and Biophysical Chemistry,
Johns Hopkins University School of Medicine
725 N. Wolfe St., Baltimore, MD 21205-2185 USA
Martin Karplus
(marci@tammy.harvard.edu)
Department of Chemistry and Chemical Biology, Harvard University
12 Oxford Street, Cambridge, MA 02138 USA
Possible collaborators
Edmund Burke
(ekb@cs.nott.ac.uk)
School of Computer Science & Information Technology
University of Nottingham
Jubilee Campus, Nottingham NG8 2BB, UK
Matthew Wilce
(mwilce@receptor.pharm.uwa.edu.au)
Department of Pharmacology/Crystallography Centre
The University of Western Australia

D.  Bibliography

1.  Candidates should show familiarity with the literature in the field.

References

[1]
C. B. Anfinsen, E. Haber, M. Sela, and F. H. White Jr. The kinetics of formation of native ribonuclease during oxidation of the reduced polypeptide chain. In Proceedings of the National Academy of Sciences of the United States of America, volume 47, pages 1309-1314, 1961.

[2]
Thomas Bäck, Ulrich Hammel, and Hans-Paul Schwefel. Evolutionary computation: Comments on the history and current state. IEEE Transactions on Evolutionary Computation, 1(1):3-16, April 1997.

[3]
L. Barone, L. While, and P. Hingston. Designing crushers with A multi-objective evolutionary algorithm. In W. B. Langdon, E. Cantú-Paz, K. Mathias, R. Roy, D. Davis, R. Poli, K. Balakrishnan, V. Honavar, G. Rudolph, J. Wegener, L. Bull, M. A. Potter, A. C. Schultz, J. F. Miller, E. Burke, and N. Jonoska, editors, GECCO 2002: Proceedings of the Genetic and Evolutionary Computation Conference, pages 995-1002, New York, July 9-13 2002. Morgan Kaufmann Publishers.

[4]
Martin J. Bayley, Gareth Jones, Peter Willett, and Michael P. Williamson. GENFOLD: A genetic algorithm for folding protein structures using NMR restraints. Protein Science, 7:491-499, 1998.

[5]
Carl Branden and John Tooze. Introduction to Protein Structure. Garland Publishing, 1991.

[6]
Bernard R. Brooks, Robert E. Bruccoleri, Barry D. Olafson, David J. States, S. Swarminathan, and Martin Karplus. CHARMM: A program for macromolecular energy, minimization, and dynamics calculations. Journal of Computational Chemistry, 4(2):187-217, 1983.

[7]
David Brown. Deciphering the message of life's assembly. Webpage, 9 April 1999.
Available: http://www.people.virginia.edu/~rjh9u/protfold.html.

[8]
B. Carr, W. Hart, N. Krasnogor, J. Hirst, E. Burke, and J. Smith. Alignment of protein structures with a memetic evolutionary algorithm. In W. B. Langdon, E. Cantú-Paz, K. Mathias, R. Roy, D. Davis, R. Poli, K. Balakrishnan, V. Honavar, G. Rudolph, J. Wegener, L. Bull, M. A. Potter, A. C. Schultz, J. F. Miller, E. Burke, and N Jonoska, editors, Proceedings of the Genetic and Evolutionary Computation Conference GECCO2002, pages 1027-1034, 9-13 July 2002.

[9]
Carlos A. Coello Coello. An updated survey of evolutionary multiobjective optimization techniques: State of the art and future trends. In Peter J. Angeline, Zbyszek Michalewicz, Marc Schoenauer, Xin Yao, and Ali Zalzala, editors, Proceedings of the Congress on Evolutionary Computation, volume 1, pages 3-13, Mayflower Hotel, Washington D.C., USA, 6-9 1999. IEEE Press.

[10]
T. E. Creighton, editor. Protein Structure: a practical approach. IRL Press at Oxford University Press, 1989.

[11]
Thomas Dandekar and Patrick Argos. Identifying the tertiary fold of small proteins with different topologies from sequence and secondary structure using the genetic algorithm and extended criteria specific for strand regions. Journal of Molecular Biology, 256:645-660, 1996.

[12]
R. Day, J. Zydallis, and G. Lamont. Solving the protein structure prediction problem through a multiobjective genetic algorithm. In Technical Proceedings of the Second International Conference on Computational Nanoscience and Nanotechnology, pages 32-35. Air Force Institute of Technology, USA, Computational Publications, April 2002.

[13]
Richard O. Day, Gary B. Lamont, and Ruth Pachter. Protein structure prediction by applying an evolutionary algorithm. In Proceedings of the Second IEE International Workshop on High Performance Computational Biology. IEEE, April 2003. Available: http://www.hicomb.org/HiCOMB2003/papers/HICOMB2003-07.pdf.

[14]
Karl R. Deerman, Gary B. Lamont, and Ruth Pachter. Linkage-learning genetic algorithm application to the protein structure prediction problem. In Proceedings of the 2001 ACM Symposium on Applied Computing, pages 333-339. ACM Press, 2001.

[15]
David B. Fogel. An introduction to evolutionary computation. Australian Journal of Intelligent Information Processing Systems, 1(2):34-42, June 1994.

[16]
David B. Fogel. Evolutionary Computation: Toward a New Philosophy of Machine Intelligence. IEEE Press, 1995.

[17]
L. J. Fogel, A. J. Owens, and M. J. Walsh. Artificial Intelligence Through Simulated Evolution. John Wiley and Sons, 1966.

[18]
Carlos M. Fonseca and Peter J. Fleming. An overview of evolutionary algorithms in multiobjective optimization. Evolutionary Computation, 3(1):1-16, 1995.

[19]
Carlos M. Fonseca and Peter J. Fleming. An overview of evolutionary algorithms in multiobjective optimization. Evolutionary Computation, 3(1):1-16, 1995.

[20]
Garrison W. Greenwood, Jae-Min Shin, Byungkook Lee, and Gary B. Fogel. A survey of recent work on evolutionary approaches to the protein folding problem. In Proceedings of the 1999 Congress on Evolutionary Computation, volume 1, pages 488-495. IEEE, 1999.

[21]
Bernd Groß, Ulrich Hammel, Peter Maldaner, Andreas Meyer, Peter Roosen, and Martin Schütz. Optimization of heat exchanger networks by means of evolution strategies. In Proceedings of the Sixth Conference on Parallel Problem Solving from Nature, pages 1002-1011. Springer, 1996.

[22]
John R. Gunn. Sampling protein conformations using segment libraries and a genetic algorithm. Journal of Chemical Physics, 106(10):4270-4281, 8 March 1997.

[23]
Frank Herrmann and Sándor Suhai. Energy minimization of peptide analogues using genetic algorithms. Journal of Computational Chemistry, 16(11):1434-1444, 1995.

[24]
J. H. Holland. Adaptation in Natural Artificial Systems. The University of Michigan Press, Ann Arbor, 1975.

[25]
K. De Jong. Evolutionary Computation: A Unified Approach. MIT Press, 2001.

[26]
Kenneth M. Merz Jr. and Scott M. Le Grand, editors. The Protein Folding Problem and Tertiary Structure Prediction. Birkhauser, 1994.

[27]
Jürgen Klockgether and Hans-Paul Schwefel. Two-phase nozzle and hollow core jet experiments. In D. G. Elliott, editor, Proceedings of the Eleventh Symposium on Engineering Aspects of Magnetohydrodynamics, pages 141-148, California Institute of Technology, Pasadena CA, March 24-26 1970.

[28]
John R. Koza. Genetic Programming: On the programming of computers by means of natural selection. MIT Press, 1992.

[29]
N. Krasnogor, B.P. Blackburne, E.K. Burke, and J.D. Hirst. Multimeme algorithms for protein structure prediction. In J. J. Merelo Guervós, P. Adamidis, H.-G. Beyer, J.-L. Fernández-Villaca nas, and H.-P. Schwefel, editors, Parallel Problem Solving from Nature VII (PPSN-2002), volume 2439 of LNCS, pages 769-778, Granada, Spain, 2002. Springer Verlag.

[30]
Natalio Krasnogor, William E. Hart, Jim Smith, and David A. Pelta. Protein structure prediction with evolutionary algorithms. In Wolfgang Banzhaf, Jason Daida, Agoston E. Eiben, Max H. Garzon, Vasant Honavar, Mark Jakiela, and Robert E. Smith, editors, Proceedings of the Genetic and Evolutionary Computation Conference, volume 2, pages 1596-1601, Orlando, Florida, USA, 13-17 1999. Morgan Kaufmann.

[31]
Natalio Krasnogor, David Pelta, Pablo E. Martinez Lopez, and Esteban de la Canal. Genetic algorithm for the protein folding problem, a critical view. In Proceedings of Engineering of Intelligent Systems. ICSC International Press, 1998.

[32]
Natalio Krasnogor and James E. Smith. Multimeme algorithms for the structure prediction and structure comparison of proteins. In Alwyn M. Barry, editor, GECCO 2002: Proceedings of the Bird of a Feather Workshops, Genetic and Evolutionary Computation Conference, pages 42-44, New York, 8 July 2002. AAAI.

[33]
Lawrence Livermore National Laboratory. Fourth community wide experiment on the critical assessment of techniques for protein structure prediction. Webpage, December 2000. Available: http://predictioncenter.llnl.gov/casp4/Casp4.html.

[34]
Natural Selection Inc. Webpage.
Available: http://www.natural-selection.com.

[35]
NuTech Software Solutions Inc. Webpage.
Available: http://www.nutech.com.

[36]
Arnold L. Patton, W. F. Punch III, and E. D. Goodman. A standard GA approach to native protein conformation problem. In Larry Eshelman, editor, Proceedings of the Sixth International Conference on Genetic Algorithms, pages 574-581. Morgan Kaufmann, 1995.

[37]
A. Piccolboni and G. Mauri. Application of evolutionary algorithms to protein folding prediction. In Proceedings of the Artificial Evolution Conference, pages 123-135, Berlin, October 1997. Springer-Verlag.

[38]
Hans-Paul Schwefel, Gunter Rudolph, and Thomas Bäck. Contemporary evolution strategies. In European Conference on Artificial Life, pages 893-907, 1995.

[39]
R. Srinivasan and G. D. Rose. LINUS: a simple algorithm to predict the fold of a protein. Proteins, 22:81-99, 1995.

[40]
Rajgopal Srinivasan and George D. Rose. Ab initio prediction of protein structure using linus. Proteins: Structure, Function, and Genetcs, 47(4):489-495, April 2002.

[41]
Michael J. E. Sternberg, editor. Protein Structure Prediction: A Practical Approach. The Practical Approach Series. Oxford University Press, 1996.

[42]
The Walking Fish Group. Webpage. Available: http://www.wfg.uwa.edu.au.

[43]
W. A. Thomasson. Unraveling the mystery of protein folding. Webpage, 22 April 2003.
Available: http://www.faseb.org/opar/protfold/protein.html.

[44]
Ron Unger and John Moult. Genetic algorithms for protein folding simulations. Journal of Molecular Biology, 231:75-81, 1993.

[45]
Jinn-Moon Yang, Chi-Hung Tsai, Ming-Jing Hwang, Huai-Kuang Tsai, Jenn-Kang Hwang, and Cheng-Yan Kao. GEM: A gaussian evolutionary method for predicting protein side-chain conformations. Protein Science, 11, 2002.

[46]
Xin Yao, editor. Evolutionary Computation: Theory and Applications. World Scientific, 1999.

[47]
Eckart Zitzler and Lothar Thiele. Multiobjective Evolutionary Algorithms: A Comparative Case Study and the Strength Pareto Approach. IEEE Transactions on Evolutionary Computation, 3(4):257-271, 1999.

E.  Facilities

1.  In addition to confirming that proper supervision is available for the project, please comment on any other requirements, for example:

Suitable supervision is available: Dr. Lyndon While has experience in the field of evolutionary computation and PhD supervision. Further expertise in the field of protein structure may need to be sought from Dr. Matthew Wilce of the Crystallography Centre at the University of Western Australia.

2.  Special Equipment - if not already available, how it will be obtained.

This research will require access to a consumer level computer terminal. The standard laboratory machines provided by the School fit this specification. However, if parallel computing architecture is needed, the Walking Fish Group [42] has funds and cluster equipment available for use.

3.  Special Techniques - may be required. If so, what are they and are expert staff available for communicating any special skills?

No special techniques are necessary for this research.

4.  Special Literature - if not available from the Library, how will access to it be obtained?

No special literature is anticipated for this project. However, if some were to arise, such literature should be available through the Library or across the Internet.

5.  Statistical Advice - is it available? If not available in the Department, how will it be obtained?

No statistical advice is necessary.

F.  Estimated Costs

1.  What funds will the department commit to maintain the project? Please give an approximate estimate of the annual amount.

All costs will be met by the School, my supervisor, or myself. No costs other than those usually associated with research are anticipated.

G.  Confidentiality & Intellectual Property

Not applicable.

H.  Approvals

No special medical or ethical approvals are required for this research.



File translated from TEX by TTH, version 3.12.
On 5 Jan 2006, 15:35.