BIOPACMAS: A Personalized, Adaptive, and Cooperative MultiAgent System for Predicting Protein Secondary Structure

In this paper, we illustrate an application aimed at predicting protein secondary structure. The proposed system has been devised using PACMAS, a generic architecture designed to support the implementation of applications explicitly tailored for
of 13
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  See discussions, stats, and author profiles for this publication at: BIOPACMAS: A Personalized, Adaptive, andCooperative MultiAgent System for PredictingProtein Secondary Structure Conference Paper   in  Lecture Notes in Computer Science · September 2005 DOI: 10.1007/11558590_59 · Source: DBLP CITATIONS 3 READS 19 5 authors , including: Some of the authors of this publication are also working on these related projects: Devising and experimenting tools for automatic or semi-automatic taxonomy generation, in supportof vertical search engines   View projectS3CLUSTER   View projectGiuliano ArmanoUniversità degli studi di Cagliari 163   PUBLICATIONS   729   CITATIONS   SEE PROFILE Massimiliano SabaUniversità degli studi di Cagliari 2   PUBLICATIONS   18   CITATIONS   SEE PROFILE All content following this page was uploaded by Giuliano Armano on 19 January 2017. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the srcinal documentand are linked to publications on ResearchGate, letting you access and read them immediately.  BIOPACMAS: A Personalized, Adaptive, andCooperative MultiAgent System for PredictingProtein Secondary Structure Giuliano Armano 1 , Gianmaria Mancosu 2 , Alessandro Orro 1 ,Massimiliano Saba 1 , and Eloisa Vargiu 1 1 University of Cagliari, Piazza d’Armi, I-09123, Cagliari, Italy { armano,orro,saba,vargiu } @diee.unica.it 2 Shardna Life Sciences, Piazza Deffenu 4, I-09121 Cagliari, Italy Abstract.  In this paper, we illustrate an application aimed at predictingprotein secondary structure. The proposed system has been devised usingPACMAS, a generic architecture designed to support the implementationof applications explicitly tailored for information retrieval tasks. PAC-MAS agents are autonomous and flexible, and can be personalized, adap-tive and cooperative depending on the given application. To investigatethe performance of the proposed approach, preliminary experiments havebeen performed on sequences taken from well-known protein databases. 1 Introduction A common problem among computer scientists, biologists, and physicians is howto share massive amounts of raw data, which are typically unstructured, dy-namic, heterogenous, and distributed over the web. In particular, accessing thewidespread amount of distributed information resources entails relevant prob-lems (e.g. “information overload” [26]). Moreover, different users are generallyinterested in different parts of the available information, so that personalizedand effective information filtering procedures are needed. From a computer sci-entist perspective, the large number of heterogeneous and dynamically changingdatabases deemed biologically relevant have drawn the attention on suitabletechnologies able to tackle the intuitive complexity of the related problems. Inour view, multiagent systems may improve the state-of-the-art in this researchfield. In fact, software agents have been widely proposed for dealing with infor-mation retrieval and filtering problems [14].From our perspective, assuming that information sources are a primary op-erational context for software agents, the following categories can be identifiedfocusing on their specific role: (i) “information agents”, able to access to infor-mation sources and to collect and manipulate such information [26], (ii) “filteragents”, able to transform information according to user preferences [25], (iii)  “task agents”, able to help users to perform tasks by solving problems and ex-changing information with other agents [16], (iv) “interface agents”, in charge of interacting with the user such that she/he interacts with other agents throughoutthem [24], and (v) “middle agents’, devised to establish communication amongrequesters and providers [11]. Although this taxonomy is focused on a quite gen-eral perspective, alternative taxonomies could be defined focusing on differentfeatures. In particular, one may focus on capabilities rather than roles, a soft-ware agent being able to embed any subset of the capabilities briefly depicted inTable 1 (together with the corresponding focus). Table 1.  Capabilities of software agents Capability Focus on the ability of ... Autonomy Operating without the intervention of users.Reactivity Reacting to a stimulus of the underlying environment ac-cording to a stimulus/response behaviour.Proactiveness Exhibiting goal-directed behavior in order to satisfy a designobjective.Social ability Interacting with other agents according to the syntax andsemantics of some selected communication language.Flexibility Exhibiting reactivity, proactiveness, and social ability simul-taneously [33].Personalization Personalizing the behavior to fulfill user’s interests and pref-erences.Adaptation Adapting to the underlying environment by learning how toreact and/or interact with it.Cooperation Interacting with other agents in order to achieve a commongoal.Deliberative capability Reasoning about the world model and of engaging planningand negotiation, possibly in coordination with other agents.Mobility Migrating from node to node in a local- or wide-area net-work. In this paper, we concentrate on the problem of predicting secondary struc-tures using a Personalized, Adaptive, and Cooperative MultiAgent System. InSection 2 relevant related work is briefly discussed. In Section 3 the Personalized,Adaptive, and Cooperative architecture, called PACMAS, is briefly depicted. InSection 4, all customizations devised for explicitly dealing with protein secondarystructure prediction are illustrated. In Section 5, preliminary experimental re-sults are briefly discussed. Section 6 draws conclusions and future work. 2 Related Work In this section, some related work is briefly recalled, according to both an ap-plicative and a technological perspective. The former is mainly focused on the  task of secondary structure prediction, whereas the latter concerns the field of software agents, which the proposed system stems from. 2.1 Secondary Protein Structure Prediction Difficulties in predicting protein structure are mainly due to the complex interac-tions between different parts of the same protein, on the one hand, and betweenthe protein and the surrounding environment, on the other hand. Actually, someconformational structures are mainly determined by local interactions betweennear residues, whereas others are due to distant interactions in the same pro-tein. Moreover, notwithstanding the fact that primary sequences are believedto contain all information necessary to determine the corresponding structure[3], recent studies demostrate that many proteins fold into their proper three-dimensional structure with the help of molecular chaperones that act as cata-lysts [15], [20]. The problem of identifying protein structures can be simplified byconsidering only their secondary structure; i.e. a linear labeling representing theconformation to which each residue belongs to. Thus, secondary structure is anabstract view of amino acid chains, in which each residue is mapped into a sec-ondary alphabet usually composed by three symbols: alpha-helix ( α ), beta-sheet( β  ), and random-coil ( c  ).A variety of secondary structure methods have been proposed in the litera-ture. Early prediction methods were based on a combination of statistical theoryand empirical rules, applied to each amino acid of the protein to be predicted [9].Artificial neural networks (ANNs) have been widely applied to this task [22] andrepresent the core of many successful secondary structure prediction methods,thanks to their ability of finding patterns without the need for predeterminedmodels or known mechanisms.The most significant innovation introduced in prediction systems was theexploitation of long-range and evolutionary information contained in multiplealignments. It is well known, in fact, that even a single variation in a sequencemay dramatically compromise its functionality. To figure out which substitutionscan possibly affect functionality, sequences that belong to the same family can bealigned, with the goal of highlighting regions that preserve a given functionality.PHD [31] is one of the first ANN-based methods that make use of evolution-ary information to perform secondary structure prediction. In particular, PHDgenerates a profile using a BLASTP [1] search; then, the result is filtered usingClustalW [21]. Furter improvement have been obtained with both more accuratemultiple alignment strategies and more powerful neural network structures. Forinstance, PSI-PRED [2] exploits the position-specific scoring matrix built duringa preprocessing performed by PSI-BLAST [23]. This approach outperforms PHDthanks to the PSI-BLAST ability of detecting distant homologies. In a more re-cent work [5], Recurrent ANNs (RANNs) are exploited to capture long-rangeinteractions. The actual system that embodies such capabilities is SSPRO [30].  2.2 Agents in Bioinformatics Many of the algorithms (pattern matching, statistical, and/or heuristic/knowledge-based) able to deal with bioinformatics issues are available to biologists in variousimplementations, and many are available over the web. Multiagent systems havea lot to contribute to these efforts. In this section, we briefly introduce twomultiagent systems that have been proposed in the literature.BIOMAS [13] is a multiagent system for automated annotation and databasestorage of sequencing data for herpesviruses. It is based on DECAF [18], a mul-tiagent system toolkit based on RETSINA [12] and TAEMS [10]. The resulting system eliminates tedious and always out-of-date hand analyses, makes the dataand annotations available for other researchers (or agent systems), and providesa level of query processing beyond even some high-profile web sites. BIOMASuses the distributed, open nature of its multiagent solution to expand the systemin several ways that will make it useful for biologists studying more organisms,and in different ways.Hermes [27] is a layered middleware system to design and execute activity-based applications in distributed environments. In particular, Hermes providesan integrated environment where application domain experts can focus on de-signing activity workflow and ignore the topological structure of the distributedenvironment. Hermes is structured as an agent-oriented system with a 3-layer ar-chitecture: run-time, system, and user. The layered middleware has been adoptedin a bioinformatics domain, where mobile agents are used to support data collec-tion and service discovery, and to simulate biological system through autonomouscomponents interactions. 3 The PACMAS Architecture PACMAS is a generic multiagent architecture, aimed at retrieving, filtering andreorganizing information according to users’ interests. PACMAS agents are al-ways autonomous and flexible; moreover, they can be personalized, adaptive,and cooperative depending on the specific application. The overall architecture(depicted in Figure 1) encompasses four main levels (i.e., information, filter, task,and interface), each of them being associated to a specific agent role. Communi-cation occurs both horizontally and vertically. The former supports cooperationamong agents belonging to a specific level. The latter supports the flow of infor-mation and/or control between adjacent levels through suitable middle-agents(which form a corresponding mid-span level).Each level of the architecture is composed by a population of agents thatcan be combined together in accordance with the following modes: pipeline,parallel, and composition. In particular, (i) pipelines can be used to distributeinformation at different levels of abstraction, so that data can be increasinglyrefined and adapted to the user’s needs, (ii) parallel connections can be usedto model a cooperation among the involved components aimed at processinginterlaced information, whereas (iii) compositions can be used for integrating
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!