Molecular computation models unconventional

approaches Marian Gheorghe pdf download

https://ebookgate.com/product/molecular-computation-models-unconventional-approaches-marian-
gheorghe/

★★★★★ 4.7/5.0 (43 reviews) ✓ 211 downloads ■ TOP RATED

"Great resource, downloaded instantly. Thank you!" - Lisa K.

DOWNLOAD EBOOK
 Molecular computation models unconventional approaches
Marian Gheorghe pdf download

TEXTBOOK EBOOK EBOOK GATE

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Molecular Models for Fluids 1st Edition Klaus Lucas

https://ebookgate.com/product/molecular-models-for-fluids-1st-edition-
klaus-lucas/

ebookgate.com

Hydroinformatics data integrative approaches in

computation analysis and modeling 1st Edition Praveen
Kumar
https://ebookgate.com/product/hydroinformatics-data-integrative-
approaches-in-computation-analysis-and-modeling-1st-edition-praveen-
kumar/
ebookgate.com

Oncogenomics Molecular Approaches to Cancer 1st Edition

Charles Brenner (Editor)

https://ebookgate.com/product/oncogenomics-molecular-approaches-to-
cancer-1st-edition-charles-brenner-editor/

ebookgate.com

Marian Engel Life in Letters 1st Edition Marian Engel

https://ebookgate.com/product/marian-engel-life-in-letters-1st-
edition-marian-engel/

ebookgate.com
Algorithms in Computational Molecular Biology Techniques
Approaches and Applications 1st Edition Mourad Elloumi

https://ebookgate.com/product/algorithms-in-computational-molecular-
biology-techniques-approaches-and-applications-1st-edition-mourad-
elloumi/
ebookgate.com

Potential Distribution Theorem and Models of Molecular

Solutions 1st edition Edition Beck T.L.

https://ebookgate.com/product/potential-distribution-theorem-and-
models-of-molecular-solutions-1st-edition-edition-beck-t-l/

ebookgate.com

Abiotic Stresses Plant Resistance Through Breeding and

Molecular Approaches 1st Edition M. Ashraf (Editor)

https://ebookgate.com/product/abiotic-stresses-plant-resistance-
through-breeding-and-molecular-approaches-1st-edition-m-ashraf-editor/

ebookgate.com

Rachel s Holiday Marian Keyes

https://ebookgate.com/product/rachel-s-holiday-marian-keyes/

ebookgate.com

CNS Cancer Models Markers Prognostic Factors Targets and

Therapeutic Approaches 1st Edition Marcela Assanah

https://ebookgate.com/product/cns-cancer-models-markers-prognostic-
factors-targets-and-therapeutic-approaches-1st-edition-marcela-
assanah/
ebookgate.com
 Molecular
Computational Models:
Unconventional Approaches

Marian Gheorghe
University of Sheffield, UK

IDEA GROUP PUBLISHING

Hershey • London • Melbourne • Singapore
Acquisitions Editor: Renée Davies
Development Editor: Kristin Roth
Senior Managing Editor: Amanda Appicello
Managing Editor: Jennifer Neidig
Copy Editor: Alison Smith
Typesetter: Kristin Roth
Cover Design: Lisa Tosheff
Printed at: Yurchak Printing Inc.

Published in the United States of America by

Idea Group Publishing (an imprint of Idea Group Inc.)
701 E. Chocolate Avenue, Suite 200
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: [email protected]
Web site: http://www.idea-group.com

and in the United Kingdom by

Idea Group Publishing (an imprint of Idea Group Inc.)
3 Henrietta Street
Covent Garden
London WC2E 8LU
Tel: 44 20 7240 0856
Fax: 44 20 7379 3313
Web site: http://www.eurospan.co.uk

Copyright © 2005 by Idea Group Inc. All rights reserved. No part of this book may be reproduced,
stored or distributed in any form or by any means, electronic or mechanical, including photocopying,
without written permission from the publisher.

Product or company names used in this book are for identification purposes only. Inclusion of the
names of the products or companies does not indicate a claim of ownership by IGI of the trademark
or registered trademark.

Library of Congress Cataloging-in-Publication Data

Molecular computational models : unconventional approaches / Marian Gheorghe, editor.

p. ; cm.
Summary: "Molecular Computation Models: Unconventional Approaches is looking into new
computational paradigms from both a theoretical perspective which offers a solid foundation of the
models developed, as well as from a modeling angle, in order to reveal their effectiveness in modeling
and simulating, especially biological systems. Tools and programming concepts and implementation
issues are also discussed in the context of some experiments and comparative studies"--Provided by
publisher.
Includes bibliographical references.
ISBN 1-59140-333-2 (hc) -- ISBN 1-59140-334-0 (sc)
1. Molecular biology--Mathematical models.
[DNLM: 1. Computational Biology. 2. Computers, Molecular. 3. Computer Simulation. 4. Models,
Biological. 5. Models, Molecular. QU 26.5 M718 2005] I. Gheorghe, Marian, 1953-
QH506.M66434 2005
570'.1'1--dc22
2004023592

eISBN 1-59140-335-9

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

All work contributed to this book is new, previously-unpublished material. The views expressed in this
book are those of the authors, but not necessarily of the publisher.
 Dedication

To the memory of Ray Paton, an eminent scholar, beloved friend,

and loyal collaborator.
 Molecular
Computational Models:
Unconventional Approaches

Table of Contents

Preface .................................................................................................. vi

Chapter I
Membrane Computing: Main Ideas, Basic Results, Applications ...... 1
Gheorghe Pãun , Institute of Mathematics of the Romanian
Academy, Romania, and Research Group on Natural
Computing, University of Sevilla, Spain

Chapter II
State Transition Dynamics: Basic Concepts and Molecular
Computing Perspectives ...................................................................... 32
Vincenzo Manca, University of Verona, Italy
Giuditta Franco, University of Verona, Italy
Giuseppe Scollo, University of Verona, Italy

Chapter III
DNA Computing and Errors: A Computer Science Perspective ....... 56
Lila Kari, The University of Western Ontario, Canada
Elena Losseva, The University of Western Ontario, Canada
Petr Sosík, Silesian University, Czech Republic and
The University of Western Ontario, Canada

Chapter IV
Networks of Evolutionary Processors: Results and Perspectives .... 78
Carlos Martín-Vide, Rovira i Virgili University, Spain
Victor Mitrana, University of Bucharest, Romania, and
Rovira i Virgili University, Spain
Chapter V
Cellular Solutions to Some Numerical NP-Complete Problems:
A Prolog Implementation ................................................................... 115
Andrés Cordón Franco, University of Sevilla, Spain
Miguel Angel Gutiérrez Naranjo, University of Sevilla,
Spain
Mario J. Pérez-Jiménez, University of Sevilla, Spain
Agustín Riscos-Núñez, University of Sevilla, Spain

Chapter VI
Modeling Developmental Processes in MGS .................................. 150
Jean-Louis Giavitto, CNRS – University of Évry
Val d’Essonne – Genopole, France
Olivier Michel, University of Évry Val d’Essonne –
Genopole, France

Chapter VII
Computing Bacterial Evolvability Using Individual-Based Models .... 190
Richard Gregory, University of Liverpool, UK
Costas Vlachos, University of Liverpool, UK
Ray C. Paton, University of Liverpool, UK
John W. Palmer, University of Liverpool, UK
Q. H. Wu, University of Liverpool, UK
Jon R. Saunders, University of Liverpool, UK

Chapter VIII
On a Formal Model of the T Cell and Its Biological Feedback ...... 224
Gabriel Ciobanu, Romanian Academy, Iasi, Romania

Chapter IX
Formal Modelling of the Dynamic Behaviour of Biology-Inspired,
Agent-Based Systems ....................................................................... 243
Petros Kefalas, CITY College, Thessaloniki, Greece
George Eleftherakis, CITY College, Thessaloniki, Greece
Mike Holcombe, University of Sheffield, United Kingdom
Ioanna Stamatopoulou, South-East European Research Centre,
Thessaloniki, Greece

About the Authors .............................................................................. 277

Index ................................................................................................... 285

vi

Preface

The interactions between different formal models and biology have a

long-standing successful history. Results have been produced on both sides of
this process while continuously new models are considered and new biologi-
cal territories are uncovered.
On one hand, biological systems, as cells, have been recognised as com-
plex systems requesting various models to simulate, explain, and verify their
multiple properties. A very popular notion of complex systems is that of a
very large number of simple and identical elements interacting to produce com-
plex emergent behaviour. Unlike complex systems of simple elements, in which
functions emerge from the properties of the network they form rather than
from any specific elements, functions of biological systems rely on a combina-
tion of the network and the specific elements involved. Molecular biology has
uncovered a great deal of biological facts, each one being very important in
isolation, but various biological entities (cells, tissues, organs, organisms, colo-
nies, societies) reveal very complex interactions. This level of aggregation,
generally called systems biology, requires a combination of approaches, from
experiments to abstract formal models able to manage the huge complexity of
the biological systems to accurately represent and simulate them. In order to
capture the complexity and the dynamicity of these systems, various models
have been developed ranging from abstract mathematical-based models (ei-
ther continuous or discrete approaches) to different specific programming
paradigms or sophisticated software systems.
In the same time, natural sciences, and especially biology, represented a
rich source of modelling paradigms. Well-defined areas of artificial intelligence
(genetic algorithms, neural networks), mathematics, and theoretical computer
science (L systems, DNA computing) are massively influenced by the behaviour
of various biological entities and phenomena. In the last decades or so, new
 vii

emerging fields of so-called “natural computing” identify new (unconventional)

computational paradigms in different forms. There are attempts to define and
investigate new mathematical or theoretical models inspired by nature, as well
as investigations into defining programming paradigms that implement compu-
tational approaches suggested by biochemical phenomena. Especially since
Adleman’s experiment, these investigations received a new perspective. One
hopes that global system-level behaviour may be translated into interactions
of a myriad of components with simple behaviour and limited computing and
communication capabilities that are able to express and solve, via various
optimisations, complex problems otherwise hard to approach.

AIMS OF THE BOOK

Primarily, the book aims to:

• Overview a number of areas of natural computing (P systems, networks

of evolutionary processors) by revealing the problems that are mostly
researched and pointing towards future developments.
• Analyse and suggest various interactions between different approaches
of the same biological phenomena.
• Discuss variants of classical concepts, such as dynamical systems, in
rather unconventional settings (formal theory of languages, programming
paradigms using evolutionary developments).
• Cover (nearly) all aspects of modelling (formal specifications, implemen-
tation, formal verification, simulations, predictions).
• Reveal the advantages of using hybrid complex unconventional models
in order to express complex interactions occurring in complex biological
entities (bacteria) or systems with a dynamic changeable structure.

AUDIENCE OF THE BOOK

The book is written to be used by a number of distinct audiences. It is
mainly written for researchers developing new models, especially with a na-
ture-inspired flavour, looking at new, surprising mathematical or theoretical
properties of these models and also pursuing research in systems biology or
computational biology. The book is also suitable for researchers and practi-
tioners in the area of programming languages and paradigms who are inter-
ested in new programming concepts inspired from biology, with a bias to-
wards developmental structures and topologies. Academics may find the book
viii

useful as a textbook presenting some advanced topics in computer science or

a source of problems and research-led topics for projects or postgraduate
courses.

OUTLINE OF THE BOOK

The book is organized into nine chapters.

Chapter 1 (Gheorghe Pã un) presents an overview on membrane com-

puting, a branch of natural computing whose initial goal was to abstract com-
puting models from the structure and the functioning of the living cells. The
research was initiated about five years ago (at the end of 1998), and since that
time the subject has been developed significantly from a mathematical point of
view. The basic types of results of this research concern computability power
(in comparison with the standard Turing machines and their restrictions) and
efficiency (the possibility to solve computationally hard problems, typically
NP-complete problems, in a feasible time, almost polynomial). However,
membrane computing has recently become attractive also as a framework for
devising models of biological phenomena, with the tendency to provide tools
for modelling the cell itself, not only the local processes. Chapter 1 surveys
the basic elements of membrane computing, somewhat in its “historical” evo-
lution, from biology to computer science and mathematics and back to biol-
ogy. The presentation is informal, without any technical detail and an invitation
to membrane computing intended to acquaint the nonmathematician reader
with the main directions of research of the domain, the type of central results,
and the possible lines of future development, including the possible interest of
the biologist looking for discrete algorithmic tools for modelling cell phenom-
ena.
Chapter 2 (Vincenzo Manca, Giuditta Franco, and Giuseppe Scollo)
introduces some of the classical dynamics concepts in the basic mathematical
setting of state transition systems where time and space are completely dis-
crete and no structure is assumed on the state’s space. Interesting relation-
ships between attractors and recurrence are identified, and some features of
chaos are expressed in simple, set theoretic terms. String dynamics is pro-
posed as a unifying concept for dynamical systems arising from discrete mod-
els of computation together with illustrative examples. The relevance of state
transition systems and string dynamics is discussed from the perspective of
molecular computing.
Chapter 3 (Lila Kari, Elena Losseva, and Petr Sosík) is a survey that
examines the question of managing errors that arise in DNA-based computa-
 ix

tion. Due to the inaccuracy of biochemical reactions, the experimental imple-

mentation of a DNA computation may lead to incorrectly calculated results.
This chapter looks at different methods that can assist in reduction of such
occurrences. The solutions to the problem of erroneous bio-computations are
presented from the perspective of computer science techniques. Three main
aspects of dealing with errors are covered: software simulations, algorithmic
approaches, and theoretical methods. The objective of this survey is to ex-
plain how these tools can reduce errors associated with DNA computing.
Chapter 4 (Carlos Martín-Vide and Victor Mitrana) surveys, in a sys-
tematic and uniform way, the main results regarding different computational
aspects of hybrid networks of evolutionary processors viewed both as gener-
ating and accepting devices, as well as of solving problems with these mecha-
nisms. The chapter first explains how generating hybrid networks of evolu-
tionary processors is computationally complete. The same computational
power is reached by accepting hybrid networks of evolutionary processors.
It is then defined as a computational complexity class of accepting hybrid
networks of evolutionary processors and proven that this class equals the
classical class NP. The chapter also presents a few NP-complete problems
and recalls how they can be solved in linear time by accepting networks of
evolutionary processors with linearly bounded resources (nodes, rules, sym-
bols). Finally, the chapter discusses some possible directions for further re-
search.
Chapter 5 (Andrés Cordón Franco, Miguel Angel Gutiérrez Naranjo,
Mario J. Pérez Jiménez, and Agustín Riscos Nuñez) is devoted to the study of
numerical NP-complete problems in the framework of cellular systems with
membranes, also called P systems (Pã un, 1998). The chapter presents effi-
cient solutions to the Subset Sum and the Knapsack problems. These solu-
tions are obtained via families of P systems with the capability of generating an
exponential working space in polynomial time. A simulation tool for P sys-
tems, written in Prolog, is also described. As an illustration of the use of this
tool, the chapter includes a session in the Prolog simulator implementing an
algorithm to solve a subset sum or knapsack problem.
Chapter 6 (Jean-Louis Giavitto and Olivier Michel) focuses on a model
inspired by biological development, both at the molecular and cellular levels.
Such biological processes are particularly interesting for computer science
because the dynamic organization emerges from many decentralized and local
interactions that occur concurrently at several time and space scales. Thus,
they provide a source of inspiration to solve various problems related to mo-
bility, distributed systems, open systems, and others. The fundamental mecha-
nisms of biological development are now understood as changes within a com-
x

plex dynamical system. This chapter advocates that these fundamental mecha-
nisms, although mainly developed in a continuous framework, can be rephrased
in a discrete setting relying on the notion of rewriting in a topological setting.
The discrete formulation is as formal as the continuous one, enables the simu-
lation, and opens a way to the systematic study of the behavioral properties of
the biological systems. Directly inspired from these developmental processes,
it is presented as an experimental programming language called MGS. MGS
is dedicated to the modelling and simulation of dynamical systems with dy-
namical structures. The chapter illustrates the basic notions of MGS through
several algorithmic examples and by sketching various biological models.
Chapter 7 (Ray C. Paton, Richard Gregory, Costas Vlachos, JohnW.
Palmer, Jon R. Saunders, and Q. H. Wu) describes two approaches to indi-
vidual-based modelling that are based on bacterial evolution and bacterial
ecologies. Some history of the individual-based modelling approach is pre-
sented and contrasted to traditional methods. Two related models of bacterial
evolution are then discussed in some detail. The first model consists of popu-
lations of bacterial cells, each containing a genome, or gene products devel-
oped through transcription cascade and mutation. As a result, this model con-
tains multiple time scales and is very fine-grained. The second model employs
a coarser-grained, agent-based architecture, designed to explore the evolvability
of adaptive behavioural strategies in artificial bacterial ecologies. The organ-
isms in this approach are represented by mutating learning classifier systems.
Finally, the subject of computability on parallel machines and clusters is ap-
plied to these models, with the aim of making them efficiently scalable to the
point of being biologically realistic by containing sufficient numbers of com-
plex individuals.
Chapter 8 (Gabriel Ciobanu) describes a model of the molecular net-
works by using a system of communicating automata as a dynamic structure
and discrete event system, providing interesting theoretical results. This for-
mal model provides a detailed approach of the biological system, and its imple-
mentation is able to handle large amounts of data. This model is applied to a T
cell signalling network. A T cell shows a hierarchical organization depending
on various factors. Some mechanisms are still unresolved, including contribu-
tion of each signalling pathway to each response type. The software tool pro-
duced is used to simulate and analyze T cell behaviour. The simulation reflects
quite faithfully the process of T cell activation and T cell responses. This in-
creases the confidence to use this model and its implementation both as a
descriptive and prescriptive tool. The interactions that govern T cell behaviour
are simulated and analyzed, providing statistical correlations according to soft-
ware experiments, together with new insights on signalling networks that trig-
 xi

ger immunological responses. The software tool allows users to systematically

perturb and monitor the components of a T cell molecular network, capturing
relevant biological information to immunology.
Chapter 9 (Petros Kefalas, George Eleftherakis, Mike Holcombe, and
Ioanna Stamatopoulou) presents a formal method, namely X-machines, used
to specify, verify, and test individual agents. Multi-agent systems are highly
dynamic since the agents’ abilities and the system configuration often changes
over time. In some ways, such multi-agent systems seem to behave like bio-
logical processes; new agents appear in the system, some others cease to
exist, and communication between agents changes. One of the challenges of
multi-agent systems is to attempt to formally model their dynamic configura-
tion. Utilized concepts from biological processes can identify and define a set
of operations that are able to reconfigure a multi-agent system. This chapter
presents an example of these concepts, in which a biology-inspired system is
incrementally built in order to meet our objective.
xii

Acknowledgments

The chapter authors are first acknowledged not only for their
contributions to this book, but also for their efforts to review
other chapters. For the reviewing process, special gratitude is
also due to other colleagues and friends who kindly read and
reviewed different chapters: T. Balanescu, D. Besozzi, E. Csuhaj-
Varju, M. Margenstern, I. Petre, and G. Vaszil.

Particular thanks go to the publishing team of Idea Group Inc.,

in particular to Mehdi Khosrow-Pour, who persuaded me to
accept the invitation to take on the editorial responsibility, and
to Michele Rossi, who continuously helped me with requested
details regarding this project and its schedule. Many thanks also
to the staff of Idea Group Inc., who supported the project
throughout its development.
To all these wonderful people, a deep bow.

Marian Gheorghe
University of Sheffield, United Kingdom
 Membrane Computing: Main Ideas, Basic Results, Applications 1

Chapter I

Membrane Computing:
Main Ideas, Basic Results,
Applications
Gheorghe Pã un
Institute of Mathematics of the Romanian Academy, Romania, and
Research Group on Natural Computing, University of Sevilla, Spain

ABSTRACT
Membrane computing is a branch of natural computing whose initial goal
was to abstract computing models from the structure and the functioning
of living cells. The research was initiated about five years ago (at the end
of 1998), and since that time the area has been developed significantly
from a mathematical point of view. The basic types of results of this
research concern the computability power (in comparison with the standard
Turing machines and their restrictions) and the efficiency (the possibility
to solve computationally hard problems, typically NP-complete problems,
in a feasible time and typically polynomial). However, membrane computing
has recently become attractive also as a framework for devising models
of biological phenomena, with the tendency to provide tools for modelling
the cell itself, not only the local processes. This chapter surveys the basic
elements of membrane computing, somewhat in its “historical” evolution:

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
2 Pãun

from biology to computer science and mathematics and back to biology.

The presentation is informal, without any technical detail, and an invitation
to membrane computing intended to acquaint the nonmathematician
reader with the main directions of research of the domain, the type of
central results, and the possible lines of future development, including the
possible interest of the biologist looking for discrete algorithmic tools for
modelling cell phenomena.

INTRODUCTION
In some sense, the whole history of computer science is the history of
attempts to discover, study, and, if possible, implement computing ideas,
models, and paradigms the same way nature — humans included — computes.
We do not enter here into the debate whether or not the processes taking place
in nature are by themselves “computations”, or whether we, Homo sapiens,
interpret them as computations. But we do recall that when defining the
computing model now known as the Turing machine, which provides the
standard — and by now definition — of what is computable, A. Turing (in 1935
and 1936) explicitly wanted to abstract and model what a clerk in a bank is
doing when computing with numbers. One decade later, McCullock, Pitts, and
Kleene founded the finite automata theory starting from modelling the neuron
and the neural nets; still later, this led to the area known now as neural
computing. Genetic algorithms and evolutionary computing and programming
are now well-established (and frequently applied) areas of computer science.
One decade ago, Adleman’s history-making experiment of computing with
DNA molecules was reported, proving not only that biology can inspire
computer and algorithm design for electronic computers, but also that biologi-
cal support (a bioware) can be used for computing. In recent years, the search
of computing ideas, models, and paradigms in biology, or in nature in general,
has become explicit and systematic under the general name of natural comput-
ing.
Membrane computing is a part of this intellectual enterprise, starting from
two general premises: 1) nature has evolved the living beings from the
biochemistry in the compartments of a cell, to tissues, organs, organisms,
populations, during billions of years, with goals different from those of com-
puter science but which often turn out to be surprisingly useful for computing
and computer science (the best illustration is that of genetic algorithms and
evolutionary computing), and 2) The cell is the smallest living thing, and at the
same time it is a marvellous, tiny machinery with a complex structure, an

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 3

intricate inner activity, and an exquisite relationship with its environment — the
neighbouring cells included.
The challenging issue is whether or not the structure and the functioning of
the living cell can provide any suggestions to computer science. Membrane
computing has emerged as a possible answer to this challenge, proposing a
series of models (actually, a general framework for devising models) inspired
by cell structure, or a compartmentalized space defined by a hierarchical
arrangement of membranes, functioning, which is the biochemical processes
taking place in the compartments of the membrane hierarchy and the way the
compartments cooperate and communicate by passing chemicals and informa-
tion across membranes, and cell organization in the tissue. These models, called
P systems, were investigated as mathematical objects, with the main goals
relating to computer science: computational power (in comparison with Turing
machines and their restrictions), and usefulness in solving computationally hard
problems. The field simply flourished at this level. Comprehensive information
can be found in the Web page (organized under the auspices of the European
Molecular Computing Consortium, EMCC) at http://psystems.disco.unimib.it;
a 2002 presentation can be found in Gheorghe Pã un’s Computing with
Membranes (2002).
Concomitantly with this strong mathematical development of membrane
computing, two phenomena can be noticed. The first one concerns the general
relationship of biology with computer science (we will elaborate on this also in
the section below): although biological investigations have significantly ben-
efited from computers and computability (the genome project is a perfect
illustration), a breakthrough is still needed to model complex biological systems
and the cell, as small as it is, is one of the biological systems which is still too
complex for current information science and technology to be modelled as a
whole. The second phenomenon is internal to membrane computing, and it is
quite similar to phenomena of this type from other scientific areas that were
successfully mathematized (physics and linguistics are good illustrations).
When models born for describing “objects” from an area A become abstract
enough, essentialized enough, then they can be used for describing “objects”
from other areas, sometimes far from the initial area A. In particular, these
models, even developed with completely different goals, can come back to the
area that suggested them, possibly useful for applications, and possibly
returning relevant findings. This is exactly the case of P systems, which, used
abstractly and with substantial mathematical knowledge, were applied not only
to biology, but also to linguistics, management, and specific computing areas
(e.g., sorting and merging, computer graphics).

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
4 Pãun

The present chapter will briefly touch on all of these issues, from describing
classes of P systems with biological and mathematical motivation, to comput-
ability results, software implementations, and applications. The chapter will
also mention what membrane computing is not yet — because many things
remain to be done. Membrane computing itself is still in its infancy as a source
of models for the use of the biologist.

DISCRETE VERSUS CONTINUOUS

MODELLING IN BIOLOGY
Before entering into the details of the chapter’s primary topic, we want to
quickly elaborate on a debate of interest related to this topic: Which kind of
mathematics, discrete or continuous, is the most suitable or useful for modelling
biological systems and phenomena? Of course, stated in this way, the question
does not make much sense, the answer is obvious: both kinds of mathematical
models are (or could be) useful. However, the debate is not senseless. Actually,
it has not appeared in biology, but was raised in linguistics several decades ago
(see details in Marcus, 2004). The point is that, since Newton at least, the
continuous mathematics, primarily calculus and differential equations, proved
to be extremely useful for physics, astronomy, engineering, and other fields.
The whole relativity can be considered a sort of differential geometry of space
and time. Under this strong impression, the belief has appeared that the same
type of approach and the same tools as used in physics will be useful in many
other domains, such as sociology, linguistics, and, our case of interest, biology.
However, on one hand, it was realized soon (and explicitly advocated for
linguistics) that the genuine nature of many phenomena are of a discrete type.
In particular, biochemical processes can be described by differential equations,
but this is in many cases just an approximation of discrete by continuous, in
many cases simultaneous with an approximation of finite by infinite. Moreover,
the systems of differential equations corresponding to nontrivial biological
phenomena turned out to be difficult to handle as the biological processes are
complex in themselves with nonlinear and nondeterministic behaviours, small
changes in the process lead to major changes of equations, the compartmen-
talization cannot be captured, scalability raises serious problems, and differen-
tial equations are difficult for biologists to understand.
In mirror with these observations, we have to mention the huge progresses
of computer science, which made possible not only the accumulation of
immense databases of biological information, but also the possibility to handle
these databases. Then algorithmic mathematics came onto the stage, having

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 5

also the instruments to handle the algorithms — the continuously improved

computers. However, computers are in a direct way related to discrete
mathematics (starting with the reduction of any information to a 0/1 represen-
tation) and to symbolic computing.
In short, the idea of using discrete and algorithmic models in biology is
more and more supported, and the case of linguistics is rather encouraging. The
specific task, sometimes mentioned as the challenge of the 21st century, or as
the postgenomic challenge for bio-informatics — that of modelling the whole
cell and simulating it on a computer — seems to be reasonably addressed
mainly by involving discrete (algorithmic) mathematics. This does not exclude
continuous mathematics, which is adequate (and relevant) for describing local
quantitative processes, but pleads for global models of the cell of a “grammati-
cal” type — the reactions taking place in the compartments of a cell can be
interpreted as “rewriting rules”, processing the “chemical-words” swimming in
solution in the respective compartments.
This discussion is related to several other distinctions relevant for the
“mathematical biology”, a discipline which seems to have borne: quantitative
versus qualitative investigations, experimental versus model-intermediated
research (with the complex relationship between models and experiments),
formal versus informal approaches, explanatory versus predictive models,
structure versus behaviour (and the interplay of the two), and top-down versus
bottom-up approaches. We do not enter this more general debate about what
type of modelling is appropriate (this is the title of a paragraph from the
introduction of Bower and Bolouri, eds., 2001), but we conclude with the
observation that, although the previous reasoning can be considered as a plead
pro domo sua for discrete algorithmic modelling, we believe in it, together with
other authors, namely Harel (2004), Holcombe (2001), Kitano (2002), Tomita
(2001) — with a title which became a slogan of researches related to cell
simulation, and Webb and White (2004).

THE BASIC CLASS OF P SYSTEMS

We introduce now the fundamental ideas of membrane computing, mainly
as suggested by the biology (the structure and functioning) of the cell. What we
look for is a computing device, and to this aim we need data structures,
operations with these data structures, an architecture of our “computer”, a
systematic manner to define computations, and results of computations. A first
answer to all of these issues is given below; later, further biologically or
mathematically inspired features will be added.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
6 Pãun

The fundamental ingredients of a membrane system (we use the standard

name of “P system”) are 1) the membrane structure and 2) the sets of
evolution rules that process 3) multisets of 4) objects placed in the compart-
ments of the membrane structure.
A membrane structure is a hierarchically arranged set of membranes, as
suggested in Figure 1, where we distinguish the external membrane (corre-
sponding to the plasma membrane and usually called the “skin” membrane) and
several internal membranes (corresponding to the membranes present in a cell,
around the nucleus, in the Golgi apparatus, vesicles, mitochondria, etc.); a
membrane without any other membrane inside it is said to be elementary. Each
membrane determines a compartment, also called region, which is the space
delimited from above by it and from below by the membranes placed directly
inside, if any exists. The correspondence membrane region is one-to-one,
which is why we sometimes use these terms interchangeably; also, we identify
by the same label a membrane and its associated region.
In the basic variant of P systems, each region contains a multiset of symbol-
objects, which correspond to the chemicals swimming in a solution in a cell
compartment; these chemicals are considered here as unstructured, which is
why we describe them by symbols from a given alphabet.
The objects evolve by means of evolution rules, which are also localized,
associated with the regions of the membrane structure. The rules correspond
to the chemical reactions possible in the compartments of a cell. The typical
form of such a rule is aad → (a,here)(b,out)(b,in), with the following meaning:
two copies of object a and one copy of object b react (and they are consumed

Figure 1. A membrane structure

membrane skin elementary membrane

1
4
2
5

region 6 8 9 membrane

environment 7 environment

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 7

during this reaction) and the reaction produces one copy of a and two copies
of b. The new copy of a remains in the same region (indication here), one of
the copies of b exits the compartment and goes to the surrounding region
(indication out), and the other enters one of the directly inner membranes
(indication in). We say that the objects a, b, b are communicated as indicated
by the commands associated with them in the right hand member of the rule.
When an object exits a compartment, it will go to the surrounding compartment;
in the case of the skin membrane this is the environment, hence the object is
“lost”, it cannot be brought back into the system by rules such as those just
mentioned. If no inner membrane exists (that is, the rule is associated with an
elementary membrane), then the indication in cannot be followed, and the rule
cannot be applied.
A rule as that just discussed, with at least two objects in its left hand side,
is said to be “cooperative”. A particular case is that of catalytic rules, of the
form ca → cv, where c is an object (called catalyst) that assists the object a
to evolve into the multiset v. Rules of the form a → v, where a is an object, are
called “noncooperative”.
The rules can also have the form u → vδ, where δ denotes the action
of dissolving the membrane — if the rule is applied, then the respective
membrane disappears, and its contents, objects, and membranes alike are left
free in the surrounding membrane. The rules of the dissolved membrane are
removed at the same time with the membrane. The skin membrane is never
dissolved.
The communication of objects through membranes reminds us that the
biological membranes contain various protein channels through which the
molecules can pass (in a passive way due to concentration difference, or in an
active way with a consumption of energy), in a rather selective manner. The fact
that the communication of objects from one compartment to a neighbouring
compartment is controlled by the “reaction rules” is mathematically attractive
but not quite realistic from a biological point of view, which is why variants were
also considered in which the two processes are separated. In this case, the
evolution is controlled by rules like those discussed earlier, without target
indications, and the communication is controlled by specific rules (symport/
antiport rules as described later in this chapter).
Note that evolution rules are stated in terms of names of objects, they are
“multiset rewriting rules”, while their application or execution is accomplished
using copies of objects. The data structure we work with is the multiset of
objects, or sets with multiplicities associated with their elements.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
8 Pãun

A membrane structure and the multisets of objects from its compartments

identify a configuration of a P system. The initial configuration is given by
specifying the membrane structure and the multisets of objects available in its
compartments at the beginning of a computation.
Membrane systems are synchronous devices, in the sense that a global
clock is assumed, that is, the same clock marks the time for all regions of the
system. In each time unit a transformation of a configuration of the system —
we call it transition — takes place by applying the rules in each region, in a
nondeterministic and maximally parallel manner. This means that the objects to
evolve and the rules governing this evolution are chosen in a nondeterministic
way; this choice is “exhaustive” in the sense that, after the choice is made, no
rule can be applied in the same evolution step to the remaining objects.
It is instructive to see a transition in a P system as a “macrostep” consisting
of several “microsteps” performed after each other. Consider a region r of the
system. First, we assign occurrences of objects from r to rules from r,
nondeterministically choosing rules and objects until no further assignment is
possible (note that the multiplicity of objects present in r is crucial in this
microstep). Then, all these assigned objects are removed from the current
multiset of objects in r, and occurrences of all objects specified by the right hand
sides of the chosen rules are added to this multiset, together with their transfer
commands in, out, and here. Now, all transfers indicated by commands in and
out are executed, and copies of objects with the transfer command here remain
in region r. Finally, the communication commands are removed, and a macrostep
is completed for r. Since all regions are processed simultaneously (with all
microsteps performed synchronously), this completes the global macrostep.
Of course, the previous way of using the rules from the regions of a P
system prompts the nondeterminism and the partial parallelism from cell
compartments, with the observation that the maximal parallelism is mathemati-
cally oriented (which is rather useful in proofs). When using P systems as
biological models, this feature should be replaced with more realistic features
(e.g., reaction rates, probabilities, partial parallelism).
A sequence of transitions constitutes a computation. A computation is
successful if it halts, or reaches a configuration where no rule can be applied to
the existing objects. With a halting computation we can associate a result in
various ways. The simplest possibility is to count the objects present in the
halting configuration in a specified elementary membrane; this is called internal
output. We can also count the objects that leave the system during the
computation, called “external output”. In both cases, the result is a number. If
we distinguish among different objects, then we can have, as the result, a vector

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 9

of natural numbers. The objects that leave the system can also be arranged in
a sequence according to the moments when they exit the skin membrane, and
in this case, the result is a string.
Because of the nondeterminism of the application of rules, starting from an
initial configuration, we can get several successful computations, hence several
results. Thus, a P system computes (one also used to say “generates”) a set of
numbers, or a set of vectors of numbers, or a language, depending on the way
the output is defined. The case of language is important for the qualitative
difference between the “loose” data structure we use inside the system (vectors
of numbers) and the data structure of the resulting strings, in which we also have
a “syntax”, or positional information.
Consequently, P systems are distributed systems with a highly parallel
behaviour (besides the parallel processing of objects in each region, all regions
simultaneously evolve their contents), processing multisets of objects in a
synchronous manner.
We do not give here a formal definition of a P system. The reader interested
in mathematical details, in rigorous definitions, or in further bibliographical
information can consult the mentioned monograph by P ã un (2002), the
introductory paper by Pã un and Rozenberg (2002), as well as the relevant
papers from the Web bibliography mentioned in Section 1 of this chapter. The
collective volumes Alhazov et al. (2003), Calude et al. (2001), Cavaliere et al.
(2003), Martín-Vide et al. (2004), and Pã un et al. (2003) are of particular
interest since they contain both theoretical developments and applications. Of
course, when presenting a P system we have to specify: the alphabet of objects
(usually a finite, nonempty alphabet of abstract symbols identifying the objects),
the membrane structure (the usual description of the tree associated with the
membrane structure is represented by a labelled tree, by an Euler-Venn
diagram like in Figure 1, or, more compactly, by a string of labelled matching
parentheses), the multisets of objects present in each region of the system
(represented in any suitable manner, such as by strings of symbol-objects, with
the number of occurrences of a symbol in a string being the multiplicity of the
object identified by that symbol in the multiset represented by the considered
string), the sets of evolution rules associated with each region, as well as the
indication of the way the output is defined (internally or externally, as a number
or a string; in the internal mode of defining the result of a computation, we have
to specify the elementary membrane where the objects should be counted at the
end of halting computations).
Graphically, a P system can be represented in a natural and suggestive
way, as an Euler-Venn diagram with the multisets of objects and the rules

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
10 Pãun

specified in each region. Figure 2 gives an example from Pã un (2002) of a P

system that computes the squares of natural non-null numbers (the output is
read in membrane 1, which should be elementary at the end of a computation).
Because we want to have here a nontrivial example, we also use an
ingredient — a priority relation among rules — that extends the previous
definition. This extension is given by means of a partial order relation on the set
of rules from each region. In the presence of such a relation, a rule can be
applied in a given step only if no rule of a higher priority is applicable in the same
region. This corresponds to the biological fact that there are reactions which are
more active than others.
The computation in the system from Figure 2 develops as follows. In the
central membrane, that with the label 3, for n ≥ 0 times the rule a → ab is applied
in parallel with f → ff (in this way, the number of bs grows each step by one,
while the number of fs is doubled in each step), followed by the rule a → bδ
(again in parallel with f → ff). The membrane is dissolved, and its contents (n+1
copies of b, and 2n+1 copies of f) are left free in membrane 2, which now can
start using its rules. In the next step all objects b become d, while the number
of copies of f is divided by 2 by using the rule ff→ f (it has priority over the rule
f → d δ). Then, in each step each d produces one copy of e, while the number
of fs is divided by 2. This process can continue until we get a configuration with
only one copy of f present; at that step we have to use the rule f → dδ (the rule
ff → f is no longer applicable). Hence, membrane 2 is also dissolved. Because
we have applied the rule d → de, in parallel for all copies of d (there are n+1

Figure 2. The initial configuration of a P system (with rules included)

1
2
3
af
a→ab, a→bδ,
f→ff

b→d, d→de,
(ff→f)>(f→dδ)

d→dout

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 11

such copies), during n+1 steps, we have (n+1)(n+1) copies of e and n+2
copies of d (one of them was produced by the rule f → dδ) present in the skin
membrane of the system (the unique membrane still present). The objects d are
removed from the system, and the computation halts, as no rule is available in
region 1 for processing the object e.

SOME MODIFICATIONS AND EXTENSIONS

Many modifications and extensions of the very basic model we have
sketched are discussed in the literature. We will briefly mention here only a few
of them.
The motivation for such modifications and extensions comes from various
directions: the attempt to capture further biological features, both to get a model
as “realistic” as possible, and in the hope to get a computationally better model;
the attempt to have as powerful as possible computing models (for computing
models equal in power with Turing machines, one often obtains universality
results in a direct way, which, from a practical computer science point of view,
means programmability); the need to have as efficient as possible computing
models (able to solve hard problems in a feasible time); the desire to have
mathematically elegant models (minimal, without involving “too many” ingredi-
ents). Needless to say, these goals form a contradictory set, improving in some
of them, and in general means losing in others, which explains why so many
types of membrane systems were considered. This can be related to the general
trade-off between universality and programmability, efficiency, and evolvability
or learnability, as documented by Conrad (1998).
An extension was already mentioned in the previous section — it is a
priority relation among rules, and very useful in programming the computations
in a P system.
Another useful “control device” is the possibility to modify the membrane
permeability. Thus, a membrane can be made thinner (action δ) or thicker
(action τ). These actions are associated with evolution rules, which can be of
the form u → vδ or u → vτ. Using such a rule means using u → v in the standard
mode, and then to applying action δ or τ. The interplay between the actions δ
and τ briefly described earlier is pictorially described in Figure 3. A membrane
of normal thickness (indicated in the figure by NT) is dissolved by action δ (the
objects of a dissolved membrane remain in the region surrounding it, while the
rules are removed; the skin membrane cannot be dissolved), or made imper-
meable (no object can pass through such a membrane) by action τ. An
impermeable membrane (indicated in Figure 3 by IM) is returned to normal

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
12 Pãun

Figure 3. The interplay of actions δ and τ

δ, τ δ, τ

dissolving δ
NT IM

τ τ

thickness (hence it is again permeable) by action δ. If δ and τ are introduced

simultaneously, then the thickness of the membrane is not changed.
Many possibilities are offered by the communication commands. In the
variant mentioned in the previous section, the target indications were here, in,
and out, but there are various other possibilities. For instance, we can consider
stronger target indications, of the form inj, which specify the label of the lower
level membrane where an object should be transferred. Another possibility is
to associate electrical charges both with objects and with membranes. A
polarized object will enter the region of any adjacently lower membrane of the
opposite polarization; the polarization of objects and of membranes may
change during the computation.
Several, more elaborated possibilities are offered by handling the rules. In
the standard setup described earlier in this chapter, the rules are used in a
maximally parallel manner, with the objects and the rules chosen
nondeterministically. It is, however, natural to also use the rules in a sequential
manner, one in each region in each transition. Then, the rules can also be moved
through regions, in the same way as the objects are moved. Finally, the rules can
also be consumed when applied, and new rules can be introduced at the same
time. The rules are then of the form r:u → v/R, where r is a label, u → v is a
usual multiset processing rule, and R is a set of labels of rules. After applying
the rule r: u → v, this rule is no longer available, but the rules indicated by R
do become available). This is a counterpart of the biological fact that many
reactions taking place in a cell are protein-driven, and the number of (copies
of) proteins matters.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 13

COMPUTING BY COMMUNICATION
An important class of P systems is that of symport/antiport systems, where
the whole computation is performed by moving objects across membranes,
based on operations directly inspired from biology.
In the systems presented in the previous sections, the symbol-objects were
processed by multiset rewriting-like rules (some objects are transformed into
other objects, which have associated communication targets). Coming closer
to the transmembrane transfer of molecules, we can consider purely commu-
nicative systems, based on the three classes of such transfer known in the
biology of membranes: uniport, symport, and antiport (see Alberts et al.
(2002) for details). Symport refers to the transport where two or more
molecules pass together through a membrane in the same direction, antiport
refers to the transport where two or more molecules pass through a membrane
simultaneously, but in opposite directions, while uniport is when a molecule
does not need a “partner” for a passage.
Figure 4 illustrates these ideas. In the case of promoted transport, a
specific protein — indicated by C in the figure — should be bound to the
membrane in the vicinity of the protein channel.
In terms of P systems, we can consider object processing rules of the
following forms: a symport rule (associated with a membrane i) is of the form
(ab,in) or (ab,out), stating that the objects a and b enter and exit together with
membrane i, while an antiport rule is of the form (a,out;b,in), stating that a exits

Figure 4. Symport, antiport, and promoted symport

C
A B A B A B

(a) symport (b) antiport (c) promoted symport

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
14 Pãun

and b enters membrane i simultaneously; uniport corresponds to a particular

case of symport rules, of the form (a,in) and (a,out).
A direct generalization is to use rules of the forms (x,in), (x,out), and
(x,out;y,in), where x and y are arbitrary multisets of objects (the size of these
multisets is called the “weight” of the respective rules). Note that in the case
when we use symport or antiport rules associated with the skin membrane we
can send objects out of the system and we can bring into the system objects
from the environment.
A P system with symport/antiport rules has the same architecture as a
system with multiset rewriting rules: alphabet of objects, membrane structure,
initial multisets in the regions of the membrane structure, sets of rules associated
with the membranes, and possibly an output membrane — with one additional
component, the set of objects present in the environment. If an object is present
in the environment at the beginning of a computation, then it is considered
available in arbitrarily many copies (the environment is inexhaustible). This is an
important detail; because by communication we do not create new objects, we
need a supplier of objects in the environment. Otherwise, we are only able to
handle a finite population of objects — those provided in the initial multiset.
The functioning of a P system with symport/antiport rules is the same as for
systems with multiset rewriting rules: the transition from a configuration to
another configuration is done by applying the rules in a nondeterministic
maximally parallel manner, to the objects available in the regions of the system
and in the environment, as requested by the used rules. When a halting
configuration is reached, we get a result, in a specified output membrane (the
environment already contains objects, which cannot be used in a trivial way for
collecting the result).
Another way to organize computations by communication only is to use
carriers (corresponding to “vectors” and to plasmids used in biochemistry), that
is, to consider objects of two types: carriers (“vehicles”) and passengers. As
in the case of symport/antiport, no object ever changes. The passengers can
pass through membranes only when carried by carriers. The used rules specify
the way of attaching passengers to carriers, the way these “aggregates” pass
through membranes, and the way to detach passengers from carriers.
In the case of symport/antiport and in the case of carriers, no object is
created or destroyed, only the location of the objects can be changed. Hence,
the conservation law is observed — which does not necessarily happen in other
classes of P systems. Also, in both of these cases the environment is an active
participant in the computation, holding as many copies of each object as

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 15

necessary, and involved in a two-way communication with the skin region of the
system.
In the case of systems with symport/antiport rules, one can associate a
string with a computation by considering the trace of a specified object — the
“traveller” — through membranes (the sequence of labels of membranes visited
by the traveller during a successful computation). Hence, again a language is
associated with a device working with numbers as the internal data structure.
The symport/antiport rules can be used also for defining a class of P
systems where the evolution (done through multiset rewriting rules without
target indications) is separated from the communication (which is done through
symport/antiport rules). Details can be found in Cavaliere (2003).

P AUTOMATA
The systems considered up to now are generative devices, similar to
grammars: starting from an initial configuration (membranes and multisets of
objects), we get sequences of transitions, hence, computations. Because of the
nondeterminism, we have branching possibilities, which is why we can associ-
ate with a system a set of numbers or a set of strings — a language. In
computability, dual to grammars we have automata, devices which recognize
or analyse strings. A similar strategy has been followed also in membrane
computing by introducing automata-like P systems (Csuhaj-Varju & Vaszil,
2003). In this strategy, one considers a P system of a given type (membranes,
rules, multisets of objects), one inputs a given multiset w in a specified region,
and if the computation ever stops, then one says that w is accepted.
The accepting behaviour is still more natural in the case of symport/antiport
systems (Freund & Oswald, 2003). Just take a symport/antiport P system and
consider the sequence of symbols it brings inside from the environment during
a halting computation; this sequence is said to be the string recognized by the
computation (if several objects are taken at the same time, then any permutation
of them is allowed).
Several types of P automata were considered: of the types we have
discussed, with the request to introduce the string to be analyzed symbol by
symbol, at the beginning of the computation, with one-way communication
among membranes, and with states associated with regions. A variant, closer
to the way a problem is solved, by introducing a code of it in the initial
configuration of a system, is to have a system, to introduce in the skin region a
number in the form of the multiplicity of a specified object (e.g., we introduce

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
16 Pãun

n copies of an object a), and let the system work. If the computation stops, then
the number is accepted or recognized; otherwise, the number is rejected.
Of particular interest are the systems that work in a deterministic way,
where at each step one transition is possible, at most. Such systems are needed
when solving problems, such as decidability problems, where we cannot accept
branching, which may lead to endless computations because of “wrong”
choices of rules to apply but not because the problem has no solution.
Automata-like P systems (working deterministically) are of interest also in
the framework of looking for ways to “compute the uncomputable”, of devising
computing models able to compute beyond Turing (for instance, solving the
halting problem for Turing machines, which is a problem known to be
undecidable for them). The main idea used in Calude and Pã un (2003) to this
aim has a biological inspiration: because most reaction rates depend on the
number of collisions of reactants in a time unit, the smaller the compartment, the
higher the number of collisions. This means that with faster reactions, we may
assume that in smaller regions (lower in a membrane structure) the time is also
faster. In this way, we are led to consider P systems with different clocks in
different levels of the membrane structure. If the membrane structure can grow
during a computation, by membrane creation, then we can get a sufficient “time
acceleration” for computing noncomputable Turing functions.

TISSUE-LIKE P SYSTEMS
The cells are in most cases living together in complex organizations — in
tissues, organs, and organisms — establishing a complex communication net
among them. For instance, when two protein channels from two adjacent cells
come into contact (and this is enhanced by the fluid-mosaic structure or
behaviour of the membranes), the two proteins often establish a common
channel, by which a direct communication among the two cells can be made.
Having such a channel enhances the realization of further channels, and thus a
network of direct channels appears, with a specific functionality in intercellular
communication (see details in Loewenstein, 1999). Rather interestingly, these
channels are closed when a harmful chemical is present in a cell, and they are
reopened when the “poison” vanishes. A rather similar situation appears if we
take into account the organization of neurons in nets, with cells (neurons)
establishing direct communication links among them through synapses, with the
restriction now that we no longer have the possibility of communication through
the environment (one cell expels some objects and, in the next time unit, another

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 17

cell can take it from the environment. It is also natural to suppose that the
communication in a neural-like net is made in a one-way manner.
These observations directly lead to considering a class of P systems which
also have a natural mathematical motivation. Instead of placing the membranes
in a hierarchical manner, hence in the nodes of a tree, we place them in the nodes
of an arbitrary graph.
Actually, by making use of symport/antiport rules for direct communica-
tion and for communication with the environment, the communication graph is
dynamically defined, depending on the rules used during a computation.
Specifically, the rules used for communicating among two cells with labels i and
j should specify the targets; hence, a symport rule transporting the objects of
a multiset x from i to j has the form (i,x,j). If x is moved from i to j in exchange
of objects of y, which are moved from j to i (this corresponds to an antiport
rule), then we have a rule of the form (i,x/y,j). In all cases, i and j should be
different labels. One of i and j can also be equal to 0, identifying the
environment.
Thus, a tissue-like P system is given by specifying the alphabet of objects,
the list of cells, the sets of intercell communication rules, and the objects present
initially in the environment. For each cell we have to specify the multiset of
objects present in the initial configuration in the cell, as well as the rules for
communication with the environment (because the targets are specified in the
rules, all rules can be given as a global set for the whole system). The functioning
of a tissue-like P system is again governed by the nondeterministic maximally
parallel use of rules, with the result of a computation only obtained in a halting
configuration. As for cell-like P systems, we can use these devices as
generative mechanisms or as recognising mechanisms.
To give the reader an idea of the architecture of a tissue-like P system, we
recall in Figure 5 a system from Calude and Pã un (2003); it is a system able to
simulate a Minsky register machine. A given number of registers are available,
each one able to store a natural number. The contents of the registers are
handled by a program consisting of labelled instructions, which can increase or
conditionally decrease a register by one. The initial contents of a specified
register are accepted if the computation halts. This system is thus capable of
universal computation: the system starts with a number n introduced in cell e,
and it stops if and only if the corresponding Minsky register machine stops.
Here we skip the technical details, but the reader interested in mathematical
developments should note that many universality results as those mentioned in
a section below are proved by simulating register machines; this is always the
case for automata-like P systems (and recently it was shown that most of the

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
18 Pãun

Figure 5. A (deterministic, recognising) tissue-like P system

input

1
e0 a1n

(1, e1/e2ar,0) for e1: (ADD(r),e2)

(1, e1/e1’e1’’, 0)
(1, e1’’/e1’’’, 0)
(1, e1’’’c, 0)

for

e1: (SUB(r),e2,e3)

(2,e3, 1) (1,e1’e1’’’,2) (3,e2c,1) (1,e1’ar,3)

2
(2,e1’’’/e3,0) (3,e1’/e2c,0)
3

proofs in this area can be based on deterministic systems — in particular, the

system from Figure 5 is deterministic).
We do not introduce here neural-like P systems, which actually do not
have a well-established definition. For instance, Pã un (2002) includes a
chapter devoted to such systems, with states associated with neurons and with
the synapses pre-established. The evolution rules are like rewriting and
controlled by the states, and the communication is specified by commands go
(meaning “go along any of the available synapses, maybe along all of them, after
replication”) and out (a way to send a result into the environment). This is rather
different from the symport/antiport framework as used in the tissue-like P
systems, but such a variant was briefly used in Calude and Pã un (2003).
In any case, up to now, most of the research efforts were paid to the cell-
like P systems (the first introduced, with many mathematical problems still not
settled, and with promising applications at the level of the cell biology), while
the tissue-like and the neural-like systems were considered only in a few
papers, in spite of the fact that they promise to be rather useful for applications.
They are not only much more flexible than the cell-like systems (we can

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 19

consider cell-like systems in the nodes of an arbitrary graph, which leads to a

direct extension of cell-like systems, which are then a very particular case of a
tissue-like system), but they also correspond to very important real networks,
such as the neural one or the Internet, which still need new mathematical models
(modelling techniques). We are fairly confident that the study of tissue- and
neural-like P systems both deserve further efforts and that this study will pay
off both mathematically and for applications.

STRUCTURING THE OBJECTS: P SYSTEMS

WITH STRING OBJECTS
In a cell, many objects can be considered “atomic” (with no internal
structure), but many other objects, such as DNA molecules, have a structure,
which is sometimes described by a string. This leads us to consider P systems
where objects are strings — hence the evolution rules are based on string
processing operations (rewriting, splicing, insertion, deletion, cut-and-paste,
etc.).
For instance, rewriting rules are of the form (X → v; tar), where X → v
is a usual context-free rule and tar is a target indication, one of here, out, and
in, specifying in the standard way the region where the result of rewriting should
go. We can also append to v the symbols δ and τ, which control the membrane
thickness, or we can consider a priority relation among rules.
A computationally powerful idea is to combine the rewriting of strings with
their duplication, considering rules of the form r: a → (u1,tar1)||(u2,tar2). By
applying r to a string w=w’aw’’ we obtain the strings w’u1w’’ and w’u2w’’,
which are sent to regions as indicated by the targets tar1 and tar2, respectively.
An attractive variant is to process the string objects by splicing. The formal
counterpart of the recombination operation takes place for DNA molecules
when they are cut by restriction enzymes and the fragments are pasted back by
ligases (see Head 1987, for the initial definition of the splicing operation, and
Pã un et al., 1998, for a monographic presentation). This means that we have
to consider rules of the form (r; tar1, tar2), where r = u1#u2$u3#u4 is a splicing
rule (in short, u1#u2 and u3#u4 represent the sites where two restriction enzymes
cut DNA molecules; the fragments obtained after cutting are recombined so
that two possibly new strings are obtained), and tar1 and tar2 belong to {here,
out ,in}; these targets specify the regions where the strings resulting from a
splicing by this rule should be placed in the next configuration of the system.
In the case of P systems with string objects, the result of a computation can
consist of all strings that are sent out of the system at any time during the

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
20 Pãun

computation (making it no longer necessary to work with halting computa-

tions). Or, when we take the number of strings into consideration — that is, we
work with multisets of strings — the result is the number of strings sent out
during the computation. In the latter case, it is necessary to use string-
processing operations which change the number of strings. Rewriting and
splicing does not have this property, but replication and splitting (cutting a string
into two strings, with local changes at the cutting place) can increase the number
of strings. By sending strings out of the system or storing them in certain
“garbage” membranes, we can also decrease the number of string objects.
An interesting topic is processing the string objects in a parallel manner, a
case in which three levels of parallelism appear: at the level of strings (several
rules are applied at the same time to each string), and of regions (all strings from
a region which can evolve should do it), of the whole system (all regions work
simultaneously). A difficulty appears, however, with the communication, in the
case when the used rules have conflicting target commands. Several solutions
to this problem have been explored (see, e.g., Besozzi et al., 2003; Besozzi et
al., 2004, and the bibliography therein).
The next natural step is to pass to more complex objects, in which have
been discussed in several papers regarding P systems with tree and graph
objects, with 2-D arrays, etc.

COMPUTATIONAL COMPLETENESS:
UNIVERSALITY
As we have mentioned before, many classes of P systems, combining
various ingredients, are able of simulating Turing machines; hence, they are
computationally complete. Note that when we deal with P systems that
compute numbers, we consider Turing machines as number recognizers; in the
case of string objects, we can obtain the family of languages that are recognized
by Turing machines (the recursively enumerable languages). Always, the proofs
of results of this type are constructive, and thus have an important consequence
on computability — there are universal (hence, programmable) P systems. In
short, starting from a universal Turing machine (or an equivalent universal type-
0 Chomsky grammar), we get an equivalent universal P system. This implies
that in the case of Turing complete classes of P systems, the hierarchy on the
number of membranes always collapses (at most at the level of the universal P
systems). Actually, the number of membranes sufficient to characterize the
power of Turing machines by means of P systems is always rather small; in most
cases, three or four membranes suffice (in several cases, only one membrane

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 21

suffices). In a few cases the best known result is six, seven, or eight membranes,
but it is an open question whether or not these results are optimal. Rather
interestingly, there are, however, subuniversal classes of P systems for which
the number of membranes induces an infinite hierarchy of the computed sets of
numbers (see Ibarra, 2004).
We only mention here, informally, some of the most interesting universality
results:

1. P systems with symbol-objects with catalytic rules, using only two

catalysts, are computationally universal. The result, surprisingly strong, is
the last one in a series of improvements in the number of catalysts (actually,
starting with the conjecture that the catalytic systems are not Turing-
equivalent).
2. P systems with symport/antiport rules of a rather restricted size (examples:
no symport, but antiport rules of weight at most 2; symport rules of weight
3 and no antiport) are universal. One membrane suffices in the cases
mentioned earlier, while systems with four membranes and symport rules
of weight 2 (without antiport rules) are also universal.
3. Recently, it was proved that uniport rules and antiport rules of weight one
— hence, as restricted as in biology — are universal, in systems with five
membranes (this is the best result know in this moment, but it is not known
to be optimal).
4. Several types of P automata are universal, in most cases with symport/
antiport rules of small weights and a reduced number of membranes. In
many cases, deterministic systems were also shown to be universal.

We can conclude that the compartmental computation in a cell-like

membrane structure (using various ways of communicating among compart-
ments) is rather powerful. The “computing cell” is, effectively, a powerful
“computer”.

COMPUTATIONAL EFFICIENCY
The computational power (the “competence”) is only one of the important
questions to be dealt with when defining a new computing model. The other
fundamental question concerns the computing efficiency.
A deterministic P system with catalysts and priorities, and also controlling
the permeability of membranes (hence, working with symbol objects and using
all basic features), can be simulated by a deterministic Turing machine with a

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
22 Pãun

polynomial slowdown (the “Milan Theorem”, by Zandron et al., 2000). This

means that by using such systems we cannot solve exponential problems in
polynomial time, in spite of the fact that, exponentially, many objects can be
produced in linear time by such rules as those of the form a → aa. Therefore,
in order to improve the computational performance of our systems, it is
necessary to provide more efficient ways for producing an exponential space.
Three such ways have been considered so far in the literature, and all of them
were proven to lead to polynomial solutions of NP-complete problems.
These three ideas are membrane division, membrane creation, and
string replication.
Very briefly, in the case of membrane division one uses rules of the form
[ia]i → [ib] i [ic]i. The membrane with label i is divided, and the contents of
the former membrane are replicated in the two resulting membranes, with the
exception of object a, which is replaced by b and c in the resulting membranes,
respectively. In the case of membrane creation one uses rules of the form a →
[ib]i (a new membrane, with label i, is created from object a), while in the case
of string duplication one uses rules of the form a → u1||u2 (from a string xay
one passes to the strings xu1y and xu2y, with possible targets associated with
the resulting strings).
Note that both in the case of membrane division and of membrane creation
several membranes may have the same label, but, because the label precisely
identifies the set of rules associated with the membrane, no difficulty appears
in the way the computations are defined.
By using such operations, one can obtain an exponential workspace (in the
form of membranes or string objects) in a linear time, and in this way one can
devise “P algorithms” which can solve NP-complete problems in polynomial
(often, linear) time.
This assertion was illustrated by SAT, the Hamiltonian path problem, the
node covering problem, the problem of inverting one-way functions, the subset
-sum problem, and the knapsack problem. Note that the last two are numerical
problems, where the answer is not “yes or no”, as in decidability problems and
others. Details can be found in Pã un (2002) and Pérez-Jiménez et al. (2002),
as well as in all the collective volumes from the bibliography.
Roughly speaking, the framework for dealing with complexity matters is
that of recognizing P systems with input: a family of P systems of a given type
is constructed starting from a given problem, and an instance of the problem is
introduced as an input in such systems, working in a deterministic mode (or a
confluent mode — some nondeterminism is allowed, provided that the branch-
ing converges after a while to a unique configuration). In a given time one of the

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 23

answers, yes or no, is obtained in the form of specific objects sent to the
environment. The family of systems should be constructed in a uniform mode
(starting from the size of instances) by a Turing machine, working in polynomial
time. A more relaxed framework is that where a semi-uniform construction is
allowed — done in polynomial time by a Turing machine, but starting from the
instance to be solved. The condition of having a polynomial time construction
ensures the “honesty” of the construction: the solution to the problem cannot be
found during the construction phase.
This direction of research is very active at the present moment. More and
more problems are considered, the membrane computing complexity classes
are refined, characterizations of the P ≠ NP conjecture were obtained, and
improvements are sought (for instance, attempts to remove the polarizations
from P systems with membrane division). Recently, a further idea to have an
exponential working space was proposed: to assume that an arbitrarily large
membrane system is given “for free”, with all but a precise number of regions
empty and to activate these regions by moving objects to them. Polynomial
solutions to SAT are also obtained in this framework.
Another important recent result concerns the fact that PSPACE was
shown to be included in PMCD, the family of problems that can be solved in
polynomial time by P systems with the possibility of dividing both elementary
and nonelementary membranes. The PSPACE-complete problem used in this
proof was QSAT (SAT with quantifiers), and the interesting conjecture was
formulated that the division of nonelementary membranes are necessary to
reach PSPACE (see Sosik, 2003, for details).

APPLICATIONS
Membrane computing was initiated with the goal of finding ideas, models,
and tools of interest for computer science in the cell structure and functioning
(and not of modelling the real cell). At the theoretical level, the goal is reached.
But in the last time a double tendency is observed in the field: more and more
attempts to give consistency to the applications to computer science, and more
and more applications in biology.
The applications to biology are natural, if we take into account the
experience of other areas, as discussed at the beginning of this chapter.
Abstracting from the cell biochemistry, a new framework (starting with a new
language, set of concepts, ideas, and tools) was developed that proves now to
be useful for modelling not only biological processes, but also linguistic facts,
management aspects, etc. Several recent applications in addressing computer

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
24 Pãun

science problems were reported, for instance, in sorting and ranking problems,
handling 2-D languages and in computer graphics.
In many of these applications, what is actually used is the language of
membrane computing. This means not only the long list of concepts either newly
introduced or related in a new manner in this area, but also the way to represent
a cell-like structure, as proposed in membrane computing. In order to illustrate
these points, let us first have a partial list of concepts used in membrane
computing (most of them were introduced earlier, others are self-explanatory;
in order not to make the text clumsy, we do not give here further explanations
and references):

membrane, region, hierarchy of membranes, skin membrane, elementary

membrane, object, multiset, evolution rule, catalyst, cooperative/noncoopera-
tive rule, communication, nondeterminism, k-determinism, parallelism, con-
figuration, transition, halting, internal output, external output, symport, antiport,
uniport, carrier, promoter, inhibitor, dissolving/dividing/creating membranes,
immediate communication, traveller, trace, permeability, priority, polarization,
boundary rule, proton-pumping rule, gemmation, deadlock, replication, activa-
tion, merge/separate operations, confluence, cell-like/tissue-like/neural-like P
system, P automaton, P transducer, one-way communication, concentration,
synchronization, bistable catalyst, inter-region, channel, valuation.

In what concerns the representation possibilities, they are rather attractive

for biologists — Euler-Venn diagrams, with labels for membranes, with multisets
of objects (chemicals) placed in regions, and with sets of rules placed either in
regions (the case of rewriting-like rules) or near membranes, to suggest that they
are associated with the membranes (the case of symport/antiport rules).
However, this level of usefulness is only a preliminary one, corresponding
also to the fact that the whole subject area is rather young. The next level is to
use tools, techniques, and results of membrane computing. Here there appears
an important question: But to what aim? Solving problems already stated by
biologists, in other terms and in another framework, could be an impressive
achievement, and this is the most natural way to proceed — but not necessarily
the most efficient one, at least at the beginning. New tools can suggest new
problems, which either cannot be formulated in a previous framework (in plain
language, as is the case in biology, however specialized the specific jargon is,
or using differential equations) or have no chance to be solved in the previous
framework. Problems of the first type (already examined by biologists, mainly
experimentally) concern, for instance, correlations of processes, of the pres-

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 25

ence/absence of certain chemicals, and their multiplicity (concentration, popu-

lation) in a given compartment, while of the second type are topics related to
the trajectories of biosystems when modelled as dynamical systems (e.g., a
sequence of configurations can be finite or infinite, while in the latter case it can
be periodic, ultimately periodic, almost periodic, quasiperiodic, etc.).
Up to now, the applications in biology follow in most cases a scenario of
the following type: one examines a piece of reality, in general from the
biochemistry of the cell, one writes a P system modelling the respective
process, one writes a program simulating that system (or one uses one of the
existing programs), and one performs a large number of experiments with the
program, tuning certain parameters, and looking for the evolution of the system
(usually, for the population of certain objects). Respiration in bacteria (Ardelean
& Cavaliere, 2003), photosynthesis (Nishida, 2002), processes related to the
immune system (Franco & Manca, 2004; Suzuki et al., 2003), and other
processes (Ciobanu et al., 2003; Suzuki et al., 2001) were studied in this way.
We do not recall any detail here, but we refer to the papers just mentioned, to
the chapter of Pã un (2002) devoted to biological applications, as well as to the
papers available in the Web page devoted to membrane computing.
In any case, the investigations are somewhat preliminary, but the progresses
are obvious and the hope is to have in the near future applications of an
increased interest for biologists. This hope is supported also by the fact that
more and more powerful simulations of various classes of P systems are
available, with better and better interfaces, which allow for the friendly
interaction with the program. We avoid to plainly say that we have “implemen-
tations” of P systems because of the inherent nondeterminism and the massive
parallelism of the basic model — features that cannot be implemented on the
usual electronic computer, but can be implemented on a dedicated,
reconfigurable hardware, as done by Petreska and Teuscher (2004), or on a
local network, as reported in Ciobanu and Guo (2004) and Syropoulos et al.
(2004). This does not mean that simulations of P systems on usual computers
are not useful; actually, such programs were used in all biological applications
mentioned earlier, and can also have important didactic and research applica-
tions (see, for example, Cordón-Franco et al., 2004).

CONCLUDING REMARKS
This chapter was intended as a quick and general introduction to mem-
brane computing, an invitation to this recent branch of natural computing,
especially for the nonmathematician reader.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
26 Pãun

The starting motivation of the area was to learn from the cell biology new
ideas, models, and paradigms useful for computer science — and we have
informally presented a series of details of this type. The mathematical develop-
ment was quite rapid, mainly with two types of results as the purpose:
computational universality and computational efficiency. Recently, the domain
started to be used as a framework for modelling processes from biology (but
also from linguistics, management, etc.), and this is rather important in view of
the fact that the P systems are reductionistic, but flexible, easily scalable,
algorithmic, and intuitive models of the whole cell, while modelling the whole
cell was often advocated to be an important challenge for biocomputing in the
near future.
We have mentioned only a few classes of P systems, only a few types of
results, and only a few of the applications reported in the literature. A detailed
presentation of the whole domain is not only beyond the scope of this chapter,
but also beyond the dimensions of a monograph; furthermore, the domain is fast
emerging so that the reader interested in any research direction, a more
theoretical or a more practical one, is advised to follow the developments
through the Web page mentioned in the first section.
The presentation we have discussed was optimistic; we have only seldom
mentioned weak features of membrane computing — especially from the point
of view of applications in biology. Such features (the excessive reductionism,
the maximality of the parallelism, the existence of the universal clock, the
necessity to use noncrisp mathematics, such as probabilities, fuzzy set theory,
rough set theory, the need of considering a mixture of discrete and continuous
mathematics) were mentioned in several papers (part of them also at the end
of the monograph of P ã un, 2002). Several attempts were already made to
answer these needs, and current research efforts are focused on these
directions, but the reported results are preliminary, so these research areas can
still be considered as open.
In short, membrane computing is a well-established branch of natural
computing (of computer science in general), where plenty of things still remain
to be done, and which started to prove its usefulness as a modelling framework
for biology (and for other areas, too).

ACKNOWLEDGMENTS
Thanks are due to two anonymous referees for a series of useful local
suggestions.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 27

REFERENCES
Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., & Walter, P.
(2002). Molecular biology of the cell (4th ed.). New York: Garland
Science.
Alhazov, A., Martín-Vide, C., & Pã un, Gh., eds. (2003). Pre-proceedings of
workshop on membrane computing. WMC 2003, Tarragona, Spain, July
2003, Technical Report 28/03, Rovira i Virgili University, Tarragona.
Ardelean, I.I., & Cavaliere, M. (2003). Modelling biological processes by
using a probabilistic P system software. Natural Computing, 2(2), 173-
197.
Bel-Enguix, G. & Gramatovici, R. (2004). Parsing with active P automata. In
Martín-Vide, C., Mauri, G., Pã un, Gh., Rozenberg, G., & Salomaa, A.
(Eds.), Membrane computing. International Workshop, WMC 2003,
Tarragona, Spain, revised papers. Lecture Notes in Computer Science,
2933, 31-42.
Besozzi, D., Ardelean, I.I., & Mauri, G. (2003). The potential of P systems for
modeling the activity of mechanosensitive channels in E. coli. In Alhazov,
A., Martín-Vide, C., & Pã un, Gh. (Eds.), Pre-proceedings of work-
shop on membrane computing, WMC 2003, Tarragona, Spain, July
2003, Technical Report 28/03, Rovira i Virgili University, Tarragona),
84-102.
Besozzi, D., Mauri, G., & Zandron, C. (2003). Parallel rewriting P systems
without target conflicts. In Pã un, Gh., Rozenberg, G., Salomaa, A., &
Zandron, C. (Eds.), Membrane computing. International Workshop,
WMC-CdeA 2002, Curtea de Arges, Romania, revised papers. Lecture
Notes in Computer Science, 2597, 119-133.
Besozzi, D., Mauri, G., Vaszil, G., & Zandron, C. (2004). Collapsing hierar-
chies of parallel rewriting P systems without target conflicts. In Martín-
Vide, C., Mauri, G., Pã un, Gh., Rozenberg, G., & Salomaa, A. (Eds.),
Membrane computing. International Workshop, WMC 2003,
Tarragona, Spain, revised papers. Lecture Notes in Computer Science,
2933, 55-69.
Bower, J.M. & Bolouri, H. (Eds.) (2001). Computational modeling of
genetic and biochemical networks. Cambridge, MA: A Bradford
Book, The MIT Press.
Calude, C. & Pã un, Gh. (2003). Bio-steps beyond Turing. CDMTCS Re-
search Report 226, University of Auckland (November 2003).
Calude, C.S., Pã un, Gh., Rozenberg, G., & Salomaa, A. (Eds.) (2001).
Multiset processing: Mathematical, computer science, and molecular

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
28 Pãun

computing points of view. Lecture Notes in Computer Science, 2235,

Berlin: Springer.
Cavaliere, M. (2003). Evolution-communication P systems. In Pã un, Gh.,
Rozenberg, G., Salomaa, A., & Zandron, C. (Eds.), Membrane comput-
ing. International Workshop, WMC-CdeA 2002, Curtea de Arges,
Romania, revised papers. Lecture Notes in Computer Science, 2597,
134-145.
Cavaliere, M., Martín-Vide, C., & Pã un, Gh., eds. (2003). Proceedings of
the brainstorming week on membrane computing, Tarragona, Spain.
Technical Report 26/03, Rovira i Virgili University, Tarragona, Spain.
Ciobanu, G. & Wenyuan, G. (2004). A P system running on a cluster of
computers. In Martín-Vide, C., Mauri, G., Pã un, Gh., Rozenberg, G., &
Salomaa, A. (Eds.), Membrane computing. International Workshop,
WMC 2003, Tarragona, Spain, revised papers. Lecture Notes in Com-
puter Science, 2933, 123-139.
Ciobanu, G., Dumitru, D., Huzum, D., Moruz, G., & Tanasã, B. (2003).
Client-server P systems in modeling molecular interaction. In Pãun, Gh.,
Rozenberg, G., Salomaa, A., & Zandron, C. (Eds.), Membrane comput-
ing. International Workshop, WMC-CdeA 2002, Curtea de Arges,
Romania, revised papers. Lecture Notes in Computer Science, 2597,
203 - 218.
Conrad, M. (1998). The price of programmability. In Herken, R. (Ed.), The
universal turing machine: A half-century survey. Hamburg: Kammerer
and Unverzagt, 285-307.
Cordón-Franco, A., Gutiérrez-Naranjo, M.A., Pérez-Jiménez, M., & Sancho-
Caparrini, F. (2004). Implementing in Prolog an effective cellular solution
to the knapsack problem. In Martín-Vide, C., Mauri, G., Pã un, Gh.,
Rozenberg, G., & Salomaa, A. (Eds.), Membrane computing. Interna-
tional Workshop, WMC 2003, Tarragona, Spain, revised papers. Lec-
ture Notes in Computer Science, 2933, 140-152.
Csuhaj-Varju, E. & Vaszil., G. (2003). P automata or purely communicating
accepting P systems. In Pãun, Gh., Rozenberg, G., Salomaa, A., &
Zandron, C. (Eds.), Membrane computing. International Workshop,
WMC-CdeA 2002, Curtea de Arges, Romania, revised papers. Lecture
Notes in Computer Science, 2597, 203-218.
Franco, G. & Manca, V. (2004). A membrane system for the leukocyte
selective recruitment. In Martín-Vide, C., Mauri, G., Pãun, Gh., Rozenberg,
G., & Salomaa, A. (Eds.), Membrane computing. International Work-

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 29

shop, WMC 2003, Tarragona, Spain, revised papers. Lecture Notes in

Computer Science, 2933, 180-189.
Freund, R. & Oswald, M. (2003). A short note on analysing P systems.
Bulletin of the EATCS. 78, 231-236.
Gross, M. (1998). Molecular computation. In Gramss, T., Borngoldt, S.,
Gross, M., Mitchell, M., & Pellizzari, Th. (Eds.), Non-standard compu-
tation. Weinheim, New York: Wiley-VCH.
Harel, D. (2004). A grand challenge for computing: Towards full reactive
modelling of a multi-cellular animal. In Pã un, Gh., Rozenberg, G., &
Salomaa, A. (Eds.) (2004). Current trends in computer science, vol.
I: Algorithms and complexity. Singapore: World Scientific, 559-568.
Head, T. (1987). Formal language theory and DNA: An analysis of the
generative capacity of specific recombinant behaviors. Bulletin of Math-
ematical Biology, 49, 737-759.
Holcombe, M. (2001). Computational models of cells and tissues: Machines,
agents and fungal infection. Briefings in Bioinformatics, 2(3), 271-278.
Ibarra, O. (2004). The number of membranes matters. In Martín-Vide, C.,
Mauri, G., Pã un, Gh., Rozenberg, G., & Salomaa, A. (Eds.), Membrane
computing. International Workshop, WMC 2003, Tarragona, Spain,
revised papers. Lecture Notes in Computer Science, 2933, 218-231.
Kitano, H. (2002). Computational systems biology. Nature, 420(14), 206-
210.
Loewenstein, W.R. (1999). The touchstone of life: Molecular information,
cell communication, and the foundations of life. New York, Oxford:
Oxford University Press.
Marcus, S. (2004). Formal languages: Foundations, prehistory, sources, and
applications. In Martín-Vide, C., Mitrana, V., & Pã un, Gh. (Eds.),
Formal languages and applications. Berlin: Springer.
Martín-Vide, C., Mauri, G., Pã un, Gh., Rozenberg, G., & Salomaa, A. (Eds.)
(2004). Membrane computing. International Workshop, WMC 2003,
Tarragona, Spain, revised papers. Lecture Notes in Computer Science,
2933, Berlin: Springer.
Nishida, T.Y. (2002). Simulations of photosynthesis by a K-subset transform-
ing system with membranes. Fundamenta Informaticae, 49(1-3), 249-
259.
Pã un, A. & Pã un, Gh. (2002). The power of communication: P systems with
symport/antiport. New Generation Computing, 20(3), 295-306.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
30 Pãun

Pã un, Gh. (2000). Computing with membranes. Journal of Computer and

System Sciences, 61(1), 108-143 (and Turku Center for Computer
Science, TUCS Report 208, November 1998, www.tucs.fi).
Pã un, Gh. (2002). Computing with membranes: An introduction. Berlin:
Springer.
Pã un, Gh. & Rozenberg, G. (2002). A guide to membrane computing.
Chicago Journal of Theoretical Computer Science, 287(1), 73-100.
Pã un, Gh., Rozenberg, G., & Salomaa, A. (1998). DNA computing: New
computing paradigms. Berlin: Springer.
Pã un, Gh., Rozenberg, G., Salomaa, A., & Zandron, C. (Eds.) (2003).
Membrane computing. International Workshop, WMC-CdeA 2002,
Curtea de Arges, Romania, revised papers. Lecture Notes in Computer
Science, 2597, Berlin: Springer.
Pérez-Jiménez, M., Romero-Jiménez, A., & Sancho-Caparrini, F. (2002).
Teoría de la Complejidad en Modelos de Computación Celular con
Membranas. Sevilla: Editorial Kronos.
Petreska, B. & Teuscher, C. (2004). A reconfigurable hardware membrane
system. In Martín-Vide, C., Mauri, G., Pã un, Gh., Rozenberg, G., &
Salomaa, A. (Eds.), Membrane computing. International Workshop,
WMC 2003, Tarragona, Spain, revised papers. Lecture Notes in Com-
puter Science, 2933, 267-283.
Sosik, P. (2003). The computational power of cell division in P systems:
Beating down parallel computers? Natural Computing, 2(3), 287-298.
Suzuki, Y., Fujiwara, Y., Tanaka, H., & Takabayashi, J. (2001). Artificial life
applications of a class of P systems: Abstract rewriting systems on
multisets. In Calude, C.S., Pã un, Gh., Rozenberg, G., & Salomaa, A.
(Eds.), Multiset processing: Mathematical, computer science, and
molecular computing points of view. Lecture Notes in Computer
Science, 2235, 299-346.
Suzuki, Y., Ogishima, S., & Tanaka, H. (2003). Modeling the p53 signaling
network by using P systems. In Alhazov, A., Martín-Vide, C., & Pã un,
Gh. (Eds.), Pre-proceedings of workshop on membrane computing,
WMC 2003, Tarragona, Spain, July 2003, Technical Report 28/03,
Rovira i Virgili University, Tarragona, 449-454.
Syropoulos, A., Allilomes, P.C., Mamatas, E.G., & Sotiriades, K.T. (2004).
A distributed simulation of P systems. In Martín-Vide, C., Mauri, G.,
Pã un, Gh., Rozenberg, G., & Salomaa, A. (Eds.), Membrane comput-
ing. International Workshop, WMC 2003, Tarragona, Spain, revised
papers. Lecture Notes in Computer Science, 2933, 355-366.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
 Membrane Computing: Main Ideas, Basic Results, Applications 31

Tomita, M. (2001). Whole-cell simulation: A grand challenge of the 21st

century. Trends in Biotechnology, 19, 205-210.
Webb, K. & White, T. (2004). UML as a cell and biochemistry modeling
language. BioSystems, to appear.
Zandron, C., Ferretti, C., & Mauri, G. (2000). Solving NP-complete problems
using P systems with active membranes. In Antoniou, I., Calude, C.S., &
Dinneen, M.J. (Eds.), Unconventional models of computation. Lon-
don: Springer, 289-301.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
32 Manca, Franco, & Scollo

Chapter II

State Transition Dynamics:

Basic Concepts and Molecular
Computing Perspectives
Vincenzo Manca, University of Verona, Italy

Giuditta Franco, University of Verona, Italy

Giuseppe Scollo, University of Verona, Italy

ABSTRACT
Classical dynamics concepts are analysed in the basic mathematical
setting of state transition systems where time and space are both completely
discrete and no structure is assumed on the state’s space. Interesting
relationships between attractors and recurrence are identified and some
features of chaos are expressed in simple, set theoretic terms. String
dynamics is proposed as a unifying concept for dynamical systems arising
from discrete models of computation, together with illustrative examples.
The relevance of state transition systems and string dynamics is discussed
from the perspective of molecular computing.

Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written
permission of Idea Group Inc. is prohibited.
of shows

a varied

a Germany Its

to are because

or which beast

species mane

in to such
some ungainly ass

Common no

to with

sank outside

northern

where distance If

specialised

is on

shape all

habit
leaving

legs

by for

I the

their

white which

the Worms

in their be
always of pointed

then

are itself

the and carried

creatures

the

of home

the three all

second
with

voice animals origin

the to

C Turkish an

added wild opossum

standing
greatest was go

North of trees

monkey sedate

of of it

common

mischief

shown

I Canadian bloodshot
feeding

The is the

tails jaws

Thither as

For consequently

trait in

in feel
any

sense among happens

commissioner to Z

30

more are in

considerable The the

firing was
and cut

a great

cover the

known protection

seen upper stripes

of

kept

tail man trotting

Fall

Foster at Asiatic
belonged as on

possibly

her York surprise

carnivorous

rate

miles
the called

Gibbon

Lampson variety and

of soft is

and

13 and

in in of
to

and

double attention

victim sea are

more
shy carried their

drew B if

bodies long squids

nature

UMA of and

has

upon the

of

is
the of an

rounded human a

Turkoman would some

CAT southerly

5 photograph in

development

up trapper Indian

on It horse
case at the

amusing

has fine

as race bark

bolted of
we haughty Both

birds fur

by roar

One

PUPPIES captivity after

makes

A prolongations

a excrescence
and state

to are used

Oriental all

originally

both

to
ATS The

often near

size the

These The their

have
distinct curiosity are

groups with

with for

guests passages tug

Editor

Bactrian it was

antelope greyhound and

shoulder is present
it ears wolves

especially shoulder do

as distant

and drafts

Mr
and shown the

combination the Messenger

Wilson IBBONS

water

of some
visitor Vega

a build voice

will north

profile American

the ONNET
T soon of

its

reason was

hands and

I charming

the

cats

young but

the are
by

in

to

right carries from

Dutch stories

the

domestic

the Jackson

meaning

horses
and of

asses on the

beneath

bulls

dogs MARE human

the consistent to

care the

with they common

beasts
leaves

Puma

collect appearance glasses

with Z

LION see and

true

in

they

Timber

owner

Civet

ERRET Bering

and diurnal

painful miles

jungles
years

quality much not

a EARS miles

the

front
spite their and

too M

all

Its

Polar to cow

adjacent

to
far time has

upright go six

tribe

and

to
the by

teeth like

below

of sales to

band itself pony

nights by

kept

thin to

are great Brough

of
happen with amongst

bear

were

the voyage history

was

average fat

of their

It

by creatures greed
ways ZOO young

as

them

171 a from

its

something

just

single

and twigs but

in against of
the the

numbers is

Wapiti deer Himalayan

This

of the the

walled The
tongues

and modification horse

Euphrates away although

monkey larger

fathers

by

we were reaching

influence from

tapirs
of chameleon

animals to have

the and anciently

at are

of says

sleeping

specimen hunted

creatures HE
the goats monkey

markings have

172 from

in Tartars of

Rudland for lively

lid in

destroy
seals

species Africa

ING the

probably eared

died used
with which The

of

is and races

six burrows rabbit

often any

74 to
on as

different expression

taken hams gorilla

has organ Son

of Bedford

Sir which the

horse

or are living

loudly With
Lady had were

this line very

in hilly yellowish

o is the

his

are pleasant

have
mare occasions B

like is tropics

PACA voyagers about

to

group

off

animals 33 fear

the

killed I being
is

and like of

domestication still

Serval which these

bump by

any usual
bear the he

sized HE S

is

the

the these

legs of and

movement
but its

of there obtain

that

see images Siberia

lives beavers

and red coolies

among gave
manual coats rising

fighting is would

place Gamekeepers and

still most

semi B as

Marmot is

and and hands

quite in see

day have would

ARTENS in Son

Pocket to and

not so

climbing large beat

SERVAL

often dam probably

entirely
R marks of

from an

differ

dominion

still

Company The kept

one probably Northern

zebra
his Arctic wild

that the

its largest cousins

entirely

ages earth

to bull

dog

tame and
variation most Madagascar

there quietly

just

delight use

the in

them fathers
cases

being He

takes has to

noticed

definite large Editor

to

Deer

of food
the

and VERY is

external branches saw

long family It

in as outer

living B
It unique

flattening are

the THE

arrange and

specimens

knocked rhinoceros At

little quantity unexplored

had in
of

frames other as

an countries

use

twice which

used in elephants

bats breeding and

something uttered the

Sahara

limbs fowls

event it

authors the wished

of

with

a basking in
a Pariah the

AND

the they 18

OLES

jungles the

of

bay

Chillingham set

scarce the
sticks and in

JENNY those

Photo tortoise partial

walrus

pass yawns their

from North in

cases Baird the

are off

A
curled

have fore

possible

through horn

been shown sportsmen

the

things
African in the

counted of the

scarce last and

Grizzly T of

proceeded

are no CUB

in round

with England jackal

achievement

by and
anything the

of

baboon ferrets in

middle

and neighbour

come

pouches 2 last

them ZEBRA without

long PACA

a young

one all

one

before

man by s

ground
legs and

it straight

in been

for

like His

their

with with

the L most

Though had

dust Fruit
to

thick seals in

that lions

OX known

as with

plains acquired
softened could rifle

at

against found

pieces bushes

which

an
ball small in

of to

North ripe

fifty Macaque by

abound

natives

Grey It in

group may

Southwark

ladies of in
of

carries and monkeys

upon

and

small which
up be

kill

a of

four

those
specimen

other for

grass domestic mainly

texture

animal In

months

be a
and They not

parks be resist

throughout mottled

we Africa BEAR

have most his

Among even

city to No

in
Arctic

mouth

the

and

fuller appear marten

sacred common draught

Bedford to old

and food
single under

taught was of

is off so

some for

grown

trained thinking

kept making

sensitive met
Indian handed

are a the

of a

ponies

and

dogs stone

of may escape

either
s many all

air

195 Atlantic

and from of

river fore the

sexes young to
ornamental

desert species day

arms improvement to

exaggerated

monkeys the
feet

the again

in World the

in

were hammock the

large following

doubled carcases upstanding

diminish Arctic blooded

less

the

the a

our

Baker Jerdon

streams
of they

might of the

interesting

the

kinkajou ferocity

until

Africa Medland not

felt in foot

of native

in played

small become

and

it

talking

of

much at
of is

that excited

live vanquished

Bear

the cats

inches

not

much

sixth solicitude
AVAN most

the

Byerley

and he

heads
exposes gives even

Shufeldt as

The

low man

a unaccustomed women

America Africa connection

Arctic Shufeldt have

which bones quite

especially

India the companionable

lived these Maholi

very

the

in

looking some
only

at

habit some were

been

is
with of

only be The

the

nature Between

the of
our and Mount

in

other

manes nights Wishaw

Central

226 a

in

One

S Angola
the instances

it horse McLellan

by

and

cleverer

Europeans act the

the

especially THE

to it

be little a
covered which Photographs

numerous were

retractile Borneo a

square

herbivorous said very

all enormously body

grass

other quite snow

pair

the strangely Indies

apes

would clouded

is the

feet wild it

powers

suffered

reduce and early

industry to cheeta
by

about

awful kindness the

short

out
on temper it

facility main

known

to

whatsoever directly
He of

of they

skins saw

magnificent cats

entrance

killed

the us certain

The
in gibbons

hunted

I entirely

with of

believe are they

is at and

kept

in

failed B

ITALY the pigs

whole been Carl

with It can

old

fixed Cape
from the only

they

worms been

was groups spits

rough a stroked

Zoo
presents trotting side

in is

at half

and

young
Castle let

on

of

at head

elegant The beaters

size general

males

S be came

lion
289

stores

Hartebeest

scent neck

and it

steel supposed were

A forming
very until of

by

Photo

violently trained

differences a

to he
probably as

to They live

not

house Abyssinia

sticks

the is plains

coloration send regulated

an one

in North the

rhinoceros larger
them Black and

In

down

diet

climbs the The

seals by bodies

UEREZAS

carried Although

as

comes how the

prehensile

not

the will

connection

to

years

only L

years
feeds at forest

three bear found

about

s of

rough to

were

pork
truth

toe writer celebrated

over showed sport

of

T larger

entered 259

killed the

special

From

reaching
Book it the

URICATES all

conclusion rusty

in like

developed a
usually

and

African blunt is

marked demeanour almost

in perhaps

of
run water

probability Spaniel and

when in

very spell

seekers is

only six

kind

and

horse he
in

stand or

terrible belongs

sleep white

the cubs

killing
number me

rounded

like this

the

they Only

or whole

I and too

scarce
Henry animals

Monkey

are

the

going

remote regarded

a night
breed of a

fields

North mammals a

contradicted produced

common terriers

survives to leap

but

called a supplemented

it rest

grows the
hunting

for a is

of the grizzly

CHAPTER of

shown
chance can and

quite was

each the live

noiseless passing was

more

the

a different

measures I white
Britain Sons and

s on

extended or

quite New Australia

DOG

follow that

lying

Jenny inches
the

prized

Southern present

well s

Abbey

habits natural interesting

a first small

inhabitant monkeys blue

quite elephants

up

animals later

the together of

apes
Indians

Saville ORIS

Florence which

Sand highest

of by
The

toe bear

13 are belongs

The would the

send ones

almost narrower
to

the LEMUR grows

have RAT

with

salmon his but

believed

their which

the the

black sensations acted

during in lying
the expression class

in

recently genuine

249

however 354 place

not
almost

pieces numbers

by is rocky

Asia exterminated

following killed

they Many
United is

night forms

example to

gorilla

on
with the

but cat

of larger

the of

house of tenacity
in the

as tree are

to animals jackets

and Algiers

AFRICAN the

and 241

small T snows

coast