Artificial Immune Systems

The classical view on the process of mutation and affinity maturation that occurs in GCs assumes that their major role is to generate high affinity levels of serum Abs, as well as a dominant pool of high affinity memory B cells, through a very efficient selection process. Here we present a model that considers different types of structures where a mutation selection process occurs, with the aim at discussing the evolution of Germinal Center reactions. Based on the results of this model, we suggest that in addition to affinity maturation, the diversity generated during the GC reaction may have also been important in the evolution towards the presently observed highly organized structure of GC in higher vertebrates.

This paper presents the computer aided simulation of a model for the control of an immune response. This model has been developed to investigate the proposed hypothesis that the same cytokine that amplifies an initiated response can eventually lead to its downregulation, if it can act on more than one cell type. The simulation environment is composed of effector cells and regulatory cells; the former, when activated, initiate an immune response, while the latter are responsible for controlling the magnitude of the response. The signalling that coordinates this process is modelled using stimulation and regulation cytokines. Simulation results obtained, in accordance with the motivating idea, are presented and discussed.

Predicting the virulence of new Influenza strains is an important problem. The solution to this problem will likely require a combination of in vitro and in silico tools that are used iteratively. We describe the agent-based modeling component of this program and report preliminary results from both the in vitro and in silico experiments.

Here we propose that frustration within dynamic interactions between cells can provide the basis for a functional immune system. Cellular frustration arises when cells in the immune system interact through exchanges of potentially conflicting and diverse signals. This results in dynamic changes in the configuration of cells that interact. If a response such as cellular activation, apoptosis or proliferation only takes place when two cells interact for a sufficiently long and characteristic time, then tolerance can be understood as the state in which no cells reach this stage and an immune response can result from a disruption of the frustrated state. Within this framework, high specificity in immune reactions is a result of a generalized kinetic proofreading mechanism that takes place at the intercellular level. An immune reaction could be directed against any cell, but this is still compatible with maintaining perfect specific tolerance against self.

We consider the human body as a well-orchestrated system of interacting swarms. Utilizing swarm intelligence techniques, we present our latest virtual simulation and experimentation environment,

IMMS:VIGO::3D

, to explore key aspects of the human immune system. Immune system cells and related entities (viruses, bacteria, cytokines) are represented as virtual agents inside 3-dimensional, decentralized and compartmentalized environments that represent primary and secondary lymphoid organs as well as vascular and lymphatic vessels. Specific immune system responses emerge as by-products from collective interactions among the involved simulated ‘agents’ and their environment. We demonstrate simulation results for clonal selection and primary and secondary collective responses after viral infection, as well as the key response patterns encountered during bacterial infection. We see this simulation environment as an essential step towards a hierarchical whole-body simulation of the immune system, both for educational and research purposes.

This paper presents an analysis of the global physical properties of an idiotypic network, using a growth model with complete dynamics. Detailed studies of the properties of idiotypic networks are valuable as one the one hand they offer a potential explanation for immunological memory, and on the other have been used by engineers in application of AIS to a range of diverse applications. The properties of both homogeneous and heterogeneous networks resulting from the model in an integer-valued shape-space are analysed and compared. In addition, the results are contrasted to those obtained using other generic growth models found in the literature which have been proposed to explain the structure and growth of biological networks, and also make a useful addition to previous published results obtained in alternative shape-spaces. We find a number of both similarities and differences with other growth models that are worthy of further study.

We investigate a minimalistic model of the idiotypic network of B-lymphocytes where idiotypes are represented by bitstrings encoding the nodes of a network. A node is occupied if a lymphocyte clone of the corresponding idiotype exists at the given moment, otherwise it is empty. There is a continuous influx of B-lymphocytes of randomly (by mutation) generated idiotype from the bone marrow. B-lymphocytes are stimulated to proliferate if its receptors (antibodies) are cross-linked by complementary structures. Unstimulated lymphocytes die. Thus, the links of the network connect nodes encoded by complementary bitstrings allowing for a few mismatches.

The random evolution leads to a network of highly organized architecture depending on only few parameters. The nodes can be classified into different groups with clearly distinct properties. We report on the building principles which allow to calculate analytically characteristics as the size and the number of links between the groups previously found by simulations.

A new specification of an immune network system is proposed. The model works on a set of antibodies from the binary shape-space and it is able to build a stable network and learn new patterns as well. A set of rules based on diversity of the repertoire of patterns which control relations of stimulation and suppression is proposed. The model is described and the results of simple experiments with the implementation of the model without and with presentation of antigens are presented.

Idiotypic network models of the immune system have long attracted interest in immunology as they offer a potential explanation for the maintenance of immunological memory. They also give a possible justification for the appearance of tolerance for a certain category of cells while maintaining immunization for the others. In this paper, we provide new evidence that the manner in which affinity is defined in an idiotypic network model imposes a definite topology on the connectivity of the potential idiotypic network that can emerge. The resulting topology is responsible for very different qualitative behaviour of the network. We show that using a 2D shape-space model with affinity based on complementary regions, a cluster-free topology results that clearly divides the space into tolerant and non-tolerant zones in which antigen are maintained or rejected respectively. On the other hand, using a 2D shape-space with an affinity function based on cell similarity, a highly clustered topology emerges in which there is no separation of the space into isolated tolerant and non-tolerant zones. Using a binary shape-space, both similar and complementary affinity measures also result in highly clustered networks. In the networks whose topologies exhibit high clustering, the tolerant and intolerant zones are so intertwined that the networks either reject all antigen or tolerate all antigen.

Permutation masks were proposed for reducing the number of holes in Hamming negative selection when applying the

r

-contiguous or

r

-chunk matching rule. Here, we show that (randomly determined) permutation masks re-arrange the semantic representation of the underlying data and therefore shatter self-regions. As a consequence, detectors do not cover areas around self regions, instead they cover randomly distributed elements across the space. In addition, we observe that the resulting holes occur in regions where actually no self regions should occur.

Gene libraries are a biological mechanism for generating combinatorial diversity in the immune system. However, they also bias the antibody creation process, so that they can be viewed as a way of guiding lifetime learning mechanisms. In this paper we examine the implications of this view, by examining coverage, avoidance of self, clustering and diversity. We show how gene libraries may improve both computational expense and performance, and present an analysis which suggests how they might do it. We suggest that gene libraries: provide combinatorial efficiency; improve coverage; reduce the cost of negative selection; and allow targeting of fixed antigen populations.

This paper motivates the use of Object Oriented technologies such as OO programming languages, UML and Design Patterns in order to facilitate the development and the communication of immune system software modeling. The introduction justifies the need for immune computer models at different levels of abstraction and for various reasons: pedagogy, testing and study of emergent phenomena and quantitative predictions. Then the benefits allowed by adopting the OO way are further illustrated by simple examples of UML class, state and sequence diagrams and instances of Design Patterns such as the “Bridge” or the “State”, helping to question and to clarify the immune objects and relationships. Finally an elementary clonal selection model, restricted to B cells, antibodies and antigens, and fully developed in the OO spirit is presented.

This paper highlights degeneracy as being an important property in both the immune system and biology in general. From this, degeneracy is chosen as a candidate to inspire artificial immune systems. As a first step in exploiting the power of degeneracy, we follow the conceptual framework approach and build an abstract computational model in order to understand the properties of degenerate detectors free of any application bias. The model we build is based on the activation of T

H

cell in the lymph node, as lymph nodes are the sites in the body where the adaptive immune response to foreign antigen in the lymph are activated. The model contains APC, antigen and T

H

cell agents that move and interact in a 2-dimensional cellular space. The T

H

cell agent receptors are assumed to be degenerate and their response to different antigen agents is measured. Initial observations and results of our model are presented and highlight some of the possibilities of degenerate detector recognition.

General properties of distance functions and of affinity functions are discussed in this paper. Reasons are given why a distance function for (n based shape-spaces should be a metric. Several distance functions that are used in shape-spaces are examined and it is shown that not all of them are metrics. It is shown which impact the type of the distance function has on the shape-space, in particular on the form of recognition or affinity regions in the shape-space. Affinity functions should be defined in such a way that they determine an affinity region with positive values inside that region and zero or negative values outside. The form of an affinity function depends on the type of the underlying distance function. This is demonstrated with several examples.

Network Intrusion Detection Systems (NIDS) monitor a network with the aim of discerning malicious from benign activity on that network. While a wide range of approaches have met varying levels of success, most IDS’s rely on having access to a database of known attack signatures which are written by security experts. Nowadays, in order to solve problems with false positive alerts, correlation algorithms are used to add additional structure to sequences of IDS alerts. However, such techniques are of no help in discovering novel attacks or variations of known attacks, something the human immune system (HIS) is capable of doing in its own specialised domain. This paper presents a novel immune algorithm for application to an intrusion detection problem. The goal is to discover packets containing novel variations of attacks covered by an existing signature base.

Self-Tolerance is a key issue in Hardware Immune Systems. Two novel detector set updating strategies are proposed in this paper as approaches to the self-tolerant problem in Hardware Immune Systems. Compared with previous detector set updating strategies, results of simulation experiments show that the detector sets being updated by the new strategies are less affected by the growing of the self set, and have a better coverage on the non-self space. Moreover, the improvement is notable when the self set is unavailable during the updating of the detector set.

Using hyperspheres as antibody recognition regions is an established abstraction which was initially proposed by theoretical immunologists for use in the modeling of antibody-antigen interactions. This abstraction is also employed in the development of many artificial immune system algorithms. Here, we show several undesirable properties of hyperspheres, especially when operating in high dimensions and discuss the problems of hyperspheres as recognition regions and how they have affected overall performance of certain algorithms in the context of real-valued negative selection.

Negative selection algorithm is one of the most important algorithms inspired by biological immune system. In this paper, a heuristic detector generation algorithm for negative selection algorithm is proposed when the partial matching rule is Hamming distance. Experimental results show that this novel detector generation algorithm has a better performance than traditional detector generation algorithm.

Artificial Immune Recognition System (AIRS) has showed an effective performance on several problems such as machine learning benchmark problems and medical classification problems like breast cancer, diabets, liver disorders classification. In this study, the resource allocation mechanism of AIRS was changed with a new one determined by Fuzzy-Logic. This system, named as Fuzzy-AIRS was used as a classifier in the diagnosis of atherosclerosis, which are of great importance in medicine. The proposed system consists of the following parts: first, we obtained features that are used as inputs for Fuzzy-AIRS from Carotid Artery Doppler Signalsusing Fast Fourier Transform (FFT), then these obtained inputs used as inputs in Fuzzy-AIRS. While AIRS algorithm obtained 75% maximum classification accuracy for 150 resources using 10-fold cross validation, Fuzzy-AIRS obtained 100% maximum classification accuracy in the same conditions. These results show that Fuzzy-AIRS proved that it could be used as an effective classifier for the medical problems.

The work explores the potentiality of a clonal selection algorithm in pattern recognition (PR). In particular, a retraining scheme for the clonal selection algorithm is formulated for better recognition of handwritten numerals (a 10-class classification problem). Empirical study with two datasets (each of which contains about 12,000 handwritten samples for 10 numerals) shows that the proposed approach exhibits very good generalization ability. Experimental results reported the average recognition accuracy of about 96%. The effect of control parameters on the performance of the algorithm is analyzed and the scope for further improvement in recognition accuracy is discussed.

The B-cell algorithm (BCA) due to Kelsey and Timmis is a function optimization algorithm inspired by the process of somatic mutation of B cell clones in the natural immune system. So far, the BCA has been shown to be perform well in comparison with genetic algorithms when applied to various benchmark optimisation problems (finding the optima of

smooth

real functions). More recently, the convergence of the BCA has been shown by Clark, Hone and Timmis, using the theory of Markov chains. However, at present the theory does not predict the average number of iterations that are needed for the algorithm to converge. In this paper we present some empirical convergence results for the BCA, using a very different

non-smooth

set of benchmark problems. We propose that certain Diophantine equations, which can be reformulated as an optimization problem in integer programming, constitute a much harder set of benchmarks for evolutionary algorithms. In the light of our empirical results, we also suggest some modifications that can be made to the BCA in order to improve its performance.

The primary objective of this paper is to put forward a general framework under which clear definitions of immune operators and their roles are provided. To this aim, a novel Population Adaptive Based Immune Algorithm (PAIA) inspired by Clonal Selection and Immune Network theories for solving multi-objective optimization problems (MOP) is proposed. The algorithm is shown to be insensitive to the initial population size; the population and clone size are adaptive with respect to the search process and the problem at hand. It is argued that the algorithm can largely reduce the number of evaluation times and is more consistent with the vertebrate immune system than the previously proposed algorithms. Preliminary results suggest that the algorithm is a valuable alternative to already established evolutionary based optimization algorithms, such as NSGA II, SPEA and VIS.

This work presents omni-aiNet, an immune-inspired algorithm developed to solve single and multi-objective optimization problems, either with single and multi-global solutions. The search engine is capable of automatically adapting the exploration of the search space according to the intrinsic demand of the optimization problem. This proposal unites the concepts of omni-optimization, already proposed in the literature, with distinctive procedures associated with immune-inspired concepts. Due to the immune inspiration, the omni-aiNet presents a population capable of adjusting its size during the execution of the algorithm, according to a predefined suppression threshold, and a new grid mechanism to control the spread of solutions in the objective space. The omni-aiNet was applied to several optimization problems and the obtained results are presented and analyzed.

The management of complex energy systems where different power sources are active in a time varying scenario of costs and prices needs efficient optimization approaches. Usually the scheduling problem is is formulated as a Mixed Integer Linear Programming (MILP) to guarantee the convergence to the global optimum. The goal of this work is to propose and compare a hybrid technique based on Artificial Immune System (AIS) and linear programming versus the traditional MILP approach. Different energy scheduling problem cases are analyzed and results of the two procedures are compared both in terms of accuracy of results and convergence speed. The work shows that, on some technical cases, AIS can efficiently tackle the energy scheduling problem in a time varying scenario and that its performances can overcome those of MILP. The obtained results are very promising and make the use of immune based procedures available for real-time management of energy systems.

Multiple sequence alignment (MSA) is one of the most important tasks in biological sequence analysis. This paper will primarily focus on on protein alignments, but most of the discussion and methodology also applies to DNA alignments. A novel hybrid clonal selection algorihm, called an aligner, is presented. It searches for a set of alignments amongst the population of candidate alignments by optimizing the classical

weighted sum of pairs

objective function. Benchmarks from BaliBASE library (v.1.0 and v.2.0) are used to validate the algorithm. Experimental results of BaliBASE v.1.0 benchmarks show that the proposed algorithm is superior to PRRP, ClustalX, SAGA, DIALIGN, PIMA, MULTIALIGN, and PILEUP8. On BaliBASE v.2.0 benchmarks the algorithm shows interesting results in terms of SP score with respect to established and leading methods, i.e. ClustalW, T-Coffee, MUSCLE, PRALINE, ProbCons, and Spem.

Intelligent Home is nowadays an established technology. Actually, most existing realizations of the Intelligent Home cannot really adapt to the needs of the inhabitants of the home so that they can learn typical user behavior. In this paper we present an AIS that can perform the usual control functions but in addition is also able to adapt to varying requirements and to learn. The AIS is network based. The antigens represent the requests to the home and the antibodies the responses to these requests. Both incorporate the relevant parameters in their structure. Antibodies are produced according to the bone marrow model and a sort of reinforcement learning mechanism is implemented. The operation of the AIS is described by a scenario.

A model for integration of low-level responses to damage, potential damage and component failure in robots is presented. This model draws on the notion of inflammation and introduces an extensible, sub-symbolic mechanism for modulating high-level behaviour using the notion of artificial inflammation. Preliminary results obtained via simulation are presented and demonstrate the potential benefits of such a scheme. Additionally the system maps the robot’s physiological state-space, which is defined in terms of the levels and sources of inflammatory response. This is achieved using Kohonen’s Self-Organizing Map algorithm to arrange the states experienced during the lifetime of the robot. The future use of this map for diagnosis and localization of faults and for the generation of specific high-level remediation behaviour is also discussed.

In this paper we outline initial concepts for an immune inspired algorithm to evaluate price time series data. The proposed solution evolves a short term pool of trackers dynamically through a process of proliferation and mutation, with each member attempting to map to trends in price movements. Successful trackers feed into a long term memory pool that can generalise across repeating trend patterns. Tests are performed to examine the algorithm’s ability to successfully identify trends in a small data set. The influence of the long term memory pool is then examined. We find the algorithm is able to identify price trends presented successfully and efficiently.

Theoretical basis of Novelty Detection in Time Series and its relationships with State Space Reconstruction are discussed. It is shown that the methods for estimation of optimal state-space reconstruction parameters may be used for the estimation of immunological novelty detection system’s parameters. This is illustrated with a V-detector system detecting novelties in Mackey-Glass time series.

There is a list of unique immune features that are currently absent from the existing artificial immune systems and other intelligent paradigms. We argue that some of AIS features can be inherent in an application itself, and thus this type of application would be a more appropriate substrate in which to develop and integrate the benefits brought by AIS. We claim here that sensor networks are such an application area, in which the ideas from AIS can be readily applied. The objective of this paper is to illustrate how closely a Danger Theory based AIS – in particular the Dendritic Cell Algorithm matches the structure and functional requirements of sensor networks. This paper also introduces a new sensor network attack called an

Interest Cache Poisoning Attack

and discusses how the DCA can be applied to detect this attack.

The Dendritic Cell algorithm (DCA) is inspired by recent work in innate immunity. In this paper a formal description of the DCA is given. The DCA is described in detail, and its use as an anomaly detector is illustrated within the context of computer security. A port scan detection task is performed to substantiate the influence of signal selection on the behaviour of the algorithm. Experimental results provide a comparison of differing input signal mappings.

Adaptive information filtering is a challenging research problem. It requires the adaptation of a representation of a user’s multiple interests to various changes in them. We investigate the application of an immune-inspired approach to this problem. Nootropia, is a user profiling model that has many properties in common with computational models of the immune system that have been based on Franscisco Varela’s work. In this paper we concentrate on Nootropia’s evaluation. We define an evaluation methodology that uses virtual user’s to simulate various interest changes. The results show that Nootropia exhibits the desirable adaptive behaviour.

We present a novel approach to incremental document maps creation, which relies upon partition of a given collection of documents into a hierarchy of homogeneous groups of documents represented by different sets of terms. Further each group (defining in fact separate context) is explored by a modified version of the aiNet immune algorithm to extract its inner structure. The immune cells produced by the algorithm become reference vectors used in preparation of the final document map. Such an approach proves to be robust in terms of time and space requirements as well as the quality of the resulting clustering model.

Spam messages are continually filling email boxes of practically every Web user. To deal with this growing problem, the development of high-performance filters to block those unsolicited messages is strongly required. An Antibody Network, more precisely SRABNET (Supervised Real-Valued Antibody Network), is proposed as an alternative filter to detect spam. The model of the antibody network is generated automatically from the training dataset and evaluated on unseen messages. We validate this approach using a public corpus, called PU1, which has a large collection of encrypted personal e-mail messages containing legitimate messages and spam. Finally, we compared the performance with the well known naïve Bayes filter using some performances indexes that will be presented.