Search

Проблем садржаности упита један је од фундаменталних проблема у рачунар- ским наукама, иницијално дефинисан за релационе упите. Са растућом популарношћу SPARQL упитног језика, проблем постаје релевантан и актуелан и у овом новом контексту. У тези је представљен оригинални приступ решавању овог проблема заснован на сво- ђењу на задовољивост у логици првог реда. Подржана је садржаност упита узимајући у обзир RDF схему, а разматра се и релација стапања, као слабија форма садржаности. Доказана је сагласност и потпуност предложеног приступа на широком спектру језич- ких конструката. Описана је и његова имплементација, у виду решавача SPECS, чији је кôд јавно доступан. Представљени су резултати детаљне експерименаталне евалуације на релевантним скуповима примера за тестирање који показују да је SPECS ефикасан, и да у поређењу са осталим савременим решавачима истог проблема даје прецизније ре- зултате у краћем времену, уз бољу покривеност језичких конструката. Једна од примена моделовања упитних језика може бити и при рефакторисању апликација које присту- пају базама података. У таквим ситуацијама, врло су честе измене којима се мењају и упити и кôд на језику у коме се они позивају. Такве промене могу сачувати укупну еквивалентност кода, док на нивоу појединачних делова еквивалентност не мора бити одржана. Коришћење алата за аутоматску верификацију еквивалентности рефактори- саног кода може да дâ гаранцију задржавања понашања програма и од суштинског је значаја за поуздан развој софтвера. Са том мотивацијом, у тези се разматра и модело- вање SQL упита у теоријама логике првог реда, којим се омогућава аутоматска провера еквивалентности C/C++ програма са уграђеним SQL-ом, што је и имплементирано у виду јавно доступног алата отвореног кода SQLAV., The query containment problem is a very important computer science problem, originally defined for relational queries. With the growing popularity of the SPARQL query language, it became relevant and important in this new context, too. This thesis introduces a new approach for solving this problem, based on a reduction to satisfiability in first order logic. The approach covers containment under RDF SCHEMA entailment regime, and it can deal with the subsumption relation, as a weaker form of containment. The thesis proves soundness and completeness of the approach for a wide range of language constructs. It also describes an implementation of the approach as an open source solver SPECS. The experimental evaluation on relevant benchmarks shows that SPECS is efficient and comparing to state-of-the-art solvers, it gives more precise results in a shorter amount of time, while supporting a larger fragment of SPARQL constructs. An application of query language modeling can be useful also along refactoring of database driven applications, where simultaneous changes that include both a query and a host language code are very common. These changes can preserve the overall equivalence, without preserving equivalence of these two parts considered separately. Because of the ability to guarantee the absence of differences in behavior between two versions of the code, tools that automatically verify code equivalence have great benefits for reliable software development. With this motivation, a custom first-order logic modeling of SQL queries is developed and described in the thesis. It enables an automated approach for reasoning about equivalence of C/C++ programs with embedded SQL. The approach is implemented within a publicly available and open source framework SQLAV.

Motivacija za izradu ove doktorske disertacije je multimedijalna kolekcija koja je rezultat vi²egodi²njih terenskih istraºivanja istraºiva£a iz Balkanolo²kog instituta Srpske akademije nauka i umetnosti. Kolekcija se sastoji od materijala u vidu snimljenih intervjua, snimljenih raznih obi£aja, pridruºenih tekstualnih opisa (protokola) i brojnih drugih dokumenata. Predmet istraºivanja ove disertacije je prou£avanje mogu¢nosti i razvoj novih metoda kojima bi se zapo£elo re²avanje problema upravljanja nematerijalnim kulturnim nasle em Balkana. Podzadaci koji se tom prilikom otvaraju su razvoj adekvatnog dizajna i implementacije multimedijalne baze podataka nematerijalnog kulturnog nasle a koja bi odgovarala potrebama razli£itih vrsta korisnika, automatska semanti£ka anotacija protokola uz pomo¢ metoda obrade prirodnih jezika, kao osnova za polu-automatsku anotaciju multimedijalne kolekcije i uspe²nu pretragu po metapodacima koji su u skladu sa CIDOC CRM standardom, istraºivanje dodatnih mogu¢nosti pretrage ove kolekcije u cilju dobijanja novih znanja, kao i razvoj izabranih metoda. Glavni problem sa dostupnim metodama je u tome ²to jo² uvek nema dovoljno razvijene infrastrukture u kontekstu obrade teksta na prirodnom jeziku, organizacije i upravljanja u domenu kulturnog nasle a na prostoru Balkana i posebno za slu£aj srpskog jezika, koja bi se mogla efektivno koristiti za re²avanje postavljenog problema. Stoga, postoji izraºena potreba za razvojem metoda kojima bi se do²lo do odgovaraju¢eg re²enja. Za polu-automatsku anotaciju multimedijalnih materijala kori²¢ena je automatska semanti£ka anotacija protokola koji su pridruºeni materijalima. Ona je sprovedena metodama ekstrakcije informacija, prepoznavanja imenovanih entiteta i ekstrakcije tema, tehnikama zasnovanim na pravilima uz pomo¢ dodatnih resursa poput elektronskih re£nika, tezaurusa i re£nika re£i iz speci£nog domena. Za klasikaciju tekstualnih protokola u odnosu na tematiku, izvedeno je istraºivanje o metodama, The motivation for writing this doctoral dissertation is a multimedia collection that is the result of many years of eld research conducted by researchers from the Institute for Balkan studies of the Serbian Academy of Sciences and Arts. The collection consists of materials in the form of recorded interviews, various recorded customs, associated textual descriptions (protocols) and numerous other documents. The subject of research of this dissertation is the study of possibilities and the development of new methods that could be used as a starting point in solving the problem of managing the intangible cultural heritage of the Balkans. The subtasks that emerge in this endeavor are the development of adequate design and implementation of a multimedia database of intangible cultural heritage that would meet the needs of dierent types of users, automatic semantic annotation of protocols using natural language processing methods, as a basis for semi-automatic annotation of the multimedia collection, and successful search by metadata which comply with the CIDOC CRM standard, study of additional search possibilities of this collection in order to gain new knowledge, as well as development of selected methods. The main problem with the available methods is that there is still not enough developed infrastructure in the context of natural language processing, organization and management in the eld of cultural heritage in the Balkans and especially for the Serbian language, which could be eectively used to solve the proposed problem. There is thus a strong need to develop methods to reach an appropriate solution. For the semi-automatic annotation of multimedia materials, automatic semantic annotation of the protocols associated with the materials was used. It was carried out by methods of information extraction, recognition of named entities and topic extraction, using rule-based techniques with the help of additional resources such as electronic dictionaries, thesauri and vo

Постоји велики број проблема који захтевају писање програмских система који имају компоненте које се извршавају међусобно асинхроно једне од других..., There is a big class of problems that require software systems with asynchronously executed components...

Struktuirano ucenje je oblast koja se bavi ucenjem složenih objekata iz složenih (struktuiranih) podataka..., Structured learning is an area that deals with the learning of complex objects from complex (structured) data. Structured data, in this context, represent data that consists of several interdependent parts. Structured learning is an important segment of machine learning, and is an active research topic with numerous unsolved problems. Applications that heavily utilize this kind of learning include textual, signal processing (audio, image and video), social, biomedical (genetic, healthcare), and many other...

Binding to Major Histocompatibility Complex (MHC) class I and class II molecules is the most restrictive step in T-cell epitope presentation. In humans these molecules are named Human Leukocyte Antigen (HLA) class I and II. However, only a minor part of MHC-binding epitopes are immunogenic. Thus, the most challinging part of T-cell prediction algorithms is to distinguish between immunogenic and nonimmunogenic epitopes. Using HLA-binding data and T-cell functional assay data we have analyzed positional biases and hydropaty of T-cell epitopes in predicted structured and unstructured regions of protein Ag, in an attempt to find out differences between T-cell immunogenic and nonimmunogenic epitopes.

Binding of peptides to MHC class I molecules is essential and the most selective step that determines T cell epitopes. Therefore, prediction of MHC-peptide binding presents the principal basis for anticipating potential T cell epitopes. The immense relevance of epitope identification in vaccine design has prompted the development of many computational methods. All of them have advantages and drawbacks. Although some available methods have reasonable accuracy, there is no guarantee that all models produce good quality predictions [1]. The aim of computational methods is to reduce the laboratory expensive experiments [2], that is way every effort to improve performance of existing methods or make reliable new method is important.

Почетак новог миленијума обележен је бурним развојем друштвених мрежа, интернет технологијама у облаку и применом вештачке интелигенције у веб алатима. Изузетно брз раст броја текстова на интернету (блогова, сајтова за електронску трговину, форума, дискусионих група, система за пренос кратких порука, друштвених мрежа и портала за објаву вести) увећао је потребу за развојем метода брзе, свеобухватне и прецизне анализе текста. Због тога је значајан развој језичких технологија чији су примарни задаци: класификација докумената (енг. Document classification), груписање докумената (енг. Document clustering), проналажење информација (енг. Information Retrieval), разрешавање значења вишезначних речи (енг. Word-sense disambiguation), екстракција из текста (енг. Text еxtraction), машинско превођење (енг. Machine translation), рачунарско препознавање говора (енг. Computer speech recognition), генерисање природног језика (енг. Natural language generation), анализа осећања (енг. sentiment analysis), итд. У рачунарској лингвистици данас је у употреби више различитих назива за област чији је предмет интересовања обрада осећања у тексту: класификација према осећању (енг. sentiment classification), истраживање мишљење (енг. opinion mining), анализа осећања (енг. sentiment analysis), екстракција осећања (енг. sentiment extraction). По својој природи и методама које користи, анализа осећања у тексту спада у област рачунарске лингвистике која се бави класификацијом текста. У процесу обраде осећања се, у општем случају, говори о три врсте класификације текстова:..., The beginning of the new millennium was marked by huge development of social networks, internet technologies in the cloud and applications of artificial intelligence tools on the web. Extremely rapid growth in the number of articles on the Internet (blogs, e-commerce websites, forums, discussion groups, and systems for transmission of short messages, social networks and portals for publishing news) has increased the need for developing methods of rapid, comprehensive and accurate analysis of the text. Therefore, remarkable development of language technologies has enabled their applying in processes of document classification, document clustering, information retrieval, word sense disambiguation, text extraction, machine translation, computer speech recognition, natural language generation, sentiment analysis, etc. In computational linguistics, several different names for the area concerning processing of emotions in text are in use: sentiment classification, opinion mining, sentiment analysis, sentiment extraction. According to the nature and the methods used, sentiment analysis in text belongs to the field of computational linguistics that deals with the classification of text. In the process of analysing of emotions we generally speak of three kinds of text classification:...

Истраживање података (eng. Data mining)је интердисциплинарно поље информатике које се бави аутоматским или полу-аутоматским откривањем знања у подацима. Основни задатак истраживања података је издвајање нетривијалних, претходно непознатих и потенцијално корисних образаца, односа и веза у подацима и статистички значајних структура из великих колекција података. Императив је да добијени резултати буду нови, ваљани, корисни и разумљиви. Технике за истраживање података укључују статистичке моделе, математичке алгоритме и методе машинског учења..., Data mining is an interdisciplinary subfield of computer science, including various scientific disciplines such as: database systems, statistics, machine learning, artificial intelligence and the others. The main task of data mining is automatic and semi-automatic analysis of large quantities of data to extract previously unknown, nontrivial and interesting patterns. Rapid development in the fields of immunology, genomics, proteomics, molecular biology and other related areas has caused a large increase in biological data. Drawing conclusions from these data requires sophisticated computational analyses. Without automatic methods to extract data it is almost impossible to investigate and analyze this data. Currently, one of the most active problems in immunoinformatics is T - cell epitope identification. Identification of T - cell epitopes, especially dominant T - cell epitopes widely represented in population, is of the immense relevance in vaccine development and detecting immunological patterns characteristic for autoimmune diseases. Epitope-based vaccines are of great importance in combating infectious and chronic diseases and various types of cancer. Experimental methods for identification of T - cell epitopes are expensive, time consuming, and are not applicable for large scale research (especially not for the choice of the optimal group of epitopes for vaccine development which will cover the whole population or personalized vaccines). Computational and mathematical models for T - cell epitope prediction, based on MHC-peptide binding, are crucial to enable the systematic investigation and identification of T - cell epitopes on a large dataset and to complement expensive and time consuming experimentation [16]. T - cells (T - lymphocytes) recognize protein antigen(s) only when degradated to peptide fragments and complexed with Major Histocompatibility Complex (MHC) molecules on the surface of antigen-presenting cells [1]. The binding of these peptides (potenti

Mnogi realni problemi se danas mogu predstaviti u obliku problema zadovoljenja ogranicenja (CSP) i zatim rešiti nekom od mnogobrojnih tehnika za rešavanje ovog problema. Jedna od tehnika podrazumeva svođenje na problem SAT, tj. problem iskazne zadovoljivosti. Promenljive i ogranicenja problema CSP se prevode (kodiraju) u SAT instancu, ona se potom rešava pomocu modernih SAT rešavaca i rešenje se, ako postoji, prevodi u rešenje problema CSP. Glavni cilj ove teze je unapređenje rešavanja problema CSP svođenjem na SAT. Razvijena su dva nova hibridna kodiranja (prevođenja u SAT formulu) koja kombinuju dobre strane postojecih kodiranja. Dat je dokaz korektnosti jednog od kodiranja koji do sada nije postojao u literaturi. Razvijen je sistem meSAT koji omogucava svođenje problema CSP na SAT pomocu cetiri osnovna i dva hibridna kodiranja, kao i rešavanje problema CSP svođenjem na dva problema srodna problemu SAT, SMT i PB. Razvijen je portfolio za automatski odabir kodiranja/rešavaca za ulaznu instancu koju je potrebno rešiti i pokazano je da je razvijeni portfolio uporediv sa najefikasnijim savremenim pristupima. Prikazan je novi pristup zasnovan na kratkim vremenskim ogranicenjima sa ciljem da se znacajno smanji vreme pripreme portfolija. Pokazano je da se ovim pristupom dobijaju rezultati konkurentni onima koji se dobijaju korišcenjem standardnog vremena za pripremu. Izvršeno je poređenje nekoliko tehnika mašinskog ucenja, sa ciljem da se ustanovi koja od njih je pogodnija za pristup sa kratkim treniranjem. Prikazan je jedan realan problem, problem raspoređivanja kontrolora leta, kao i tri njegova modela. Veliki broj metoda rešavanja i raznovrsnih rešavaca je upotrebljeno za rešavanje ovog problema. Razvijeno je više optimizacionih tehnika koje imaju za cilj pronalaženje optimalnih rešenja problema. Pokazuje se da je najefikasnija hibridna tehnika koja kombinuje svođenje na SAT i lokalnu pretragu. Razmotren je i problem sudoku, kao i postojece tehnike rešavanja sudoku za, Many real-world problems can be modeled as constraint satisfaction problems (CSPs) and then solved by one of many available techniques for solving these problems. One of the techniques is reduction to SAT, i.e. Boolean Satisfiability Problem. Variables and constraints of CSP are translated (encoded) to SAT instance, that is then solved by state-of-the-art SAT solvers and solution, if exists, is translated to the solution of the original CSP. The main aim of this thesis is to improve CSP solving techniques that are using reduction to SAT. Two new hybrid encodings of CSPs to SAT are presented and they combine good sides of the existing encodings. We give the proof of correctness of one encoding that did not exist in literature. We developed system meSAT that enables reduction of CSPs to SAT by using 4 basic and 2 hybrid encodings. The system also enables solving of CSPs by reduction to two problems related to SAT, SMT and PB. We developed a portfolio for automated selection of encoding/solver to be used on some new instance that needs to be solved. The developed portfolio is comparable with the state-of-the-art portfolios. We developed a hybrid approach based on short solving timeouts with the aim of significantly reducing the preparation time of a portfolio. By using this approach, we got results comparable to the ones obtained by using preparation time of usual length. We made comparison between several machine learning techniques with the aim to find out which one is the best suited for the short training approach. The problem of assigning air traffic controllers to shifts is described and three models of this problem are presented. We used a large number of different solving methods and a diverse set of solvers for solving this problem. We developed optimization techniques that aim to find optimal solutions of the problem. A hybrid technique combining reduction to SAT and local search is shown to be the most efficient one. We also considered sudoku puzzles and the

We have developed EpDis and MassPred, extendable open source software tools that support bioinformatic research and enable parallel use of different methods for the prediction of T cell epitopes, disorder and disordered binding regions and hydropathy calculation. These tools offer a semi-automated installation of chosen sets of external predictors and an interface allowing for easy application of the prediction methods, which can be applied either to individual proteins or to datasets of a large number of proteins. In addition to access to prediction methods, the tools also provide visualization of the obtained results, calculation of consensus from results of different methods, as well as import of experimental data and their comparison with results obtained with different predictors. The tools also offer a graphical user interface and the possibility to store data and the results obtained using all of the integrated methods in the relational database or flat file for further analysis. The MassPred part enables a massive parallel application of all integrated predictors to the set of proteins. Both tools can be downloaded from http://bioinfo.matf.bg.ac.rs/home/downloads.wafl?cat=Software. Appendix A includes the technical description of the created tools and a list of supported predictors.

Istraživanje podataka (eng. Data mining)je interdisciplinarno polje informatike koje se bavi automatskim ili polu-automatskim otkrivanjem znanja u podacima. Osnovni zadatak istraživanja podataka je izdvajanje netrivijalnih, prethodno nepoznatih i potencijalno korisnih obrazaca, odnosa i veza u podacima i statistički značajnih struktura iz velikih kolekcija podataka. Imperativ je da dobijeni rezultati budu novi, valjani, korisni i razumljivi. Tehnike za istraživanje podataka uključuju statističke modele, matematičke algoritme i metode mašinskog učenja..., Data mining is an interdisciplinary subfield of computer science, including various scientific disciplines such as: database systems, statistics, machine learning, artificial intelligence and the others. The main task of data mining is automatic and semi-automatic analysis of large quantities of data to extract previously unknown, nontrivial and interesting patterns. Rapid development in the fields of immunology, genomics, proteomics, molecular biology and other related areas has caused a large increase in biological data. Drawing conclusions from these data requires sophisticated computational analyses. Without automatic methods to extract data it is almost impossible to investigate and analyze this data. Currently, one of the most active problems in immunoinformatics is T - cell epitope identification. Identification of T - cell epitopes, especially dominant T - cell epitopes widely represented in population, is of the immense relevance in vaccine development and detecting immunological patterns characteristic for autoimmune diseases. Epitope-based vaccines are of great importance in combating infectious and chronic diseases and various types of cancer. Experimental methods for identification of T - cell epitopes are expensive, time consuming, and are not applicable for large scale research (especially not for the choice of the optimal group of epitopes for vaccine development which will cover the whole population or personalized vaccines). Computational and mathematical models for T - cell epitope prediction, based on MHC-peptide binding, are crucial to enable the systematic investigation and identification of T - cell epitopes on a large dataset and to complement expensive and time consuming experimentation [16]. T - cells (T - lymphocytes) recognize protein antigen(s) only when degradated to peptide fragments and complexed with Major Histocompatibility Complex (MHC) molecules on the surface of antigen-presenting cells [1]. The binding of these peptides (potenti

У овом раду предложен је приступ за убрзавање откривања употребе софтверске библиотеке која је доступна у облику изворног кода и која је бесправно уграђена у бинарни код. Претпоставка је да су из бинарног кода одстрањене све додатне информације које би могле помоћи приликом откривања употребе софтверске библиотеке, што је чест случај у наменским уређајима. Стога је у раду коришћена ARM архитектура која је једна од најзаступљенијих у наменским уређајима. Приступ је заснован на софтверским метрикама и састоји се од две фазе. У првој фази рачуна се сличност између бинарног облика тражене процедуре (процедура из библиотеке) и сваке од процедура бинарног кода за који се сумња да користи библиотеку. У другој фази се рангирају процедуре из посматраног бинарног кода у складу са израчунатом сличношћу према траженој процедури. На крају се издваја првих N процедура и даље анализира ручно, при чему експерт бира вредност N у складу са расположивим временом и важношћу откривања неовлашћене употребе библиотеке. Одређивање сличности између две бинарне процедуре почиње прикупљањем вредности софтверских метрика. У раду је предложено укупно 19 различитих софтверских метрика. Вредности софтверских метрика..., This work proposes an approach to accelerate the discovery of a case when a software library, available in the source code, is used in a binary code without an appropriate permission. It is assumed that the binary code does not contain any additional information that could help in detecting the use of the software library, which is often the case in embedded devices. This study uses ARM architecture which is one of the most commonly found architectures in embedded devices. Approach is based on software metrics and consists of two phases. The first phase calculates similarity between the searched binary procedure originating from the library, and each of the procedures found in the binary code that is suspected to use the library. In the second phase of the proposed approach, the procedures from the binary code are ranked in accordance with the similarity to the searched procedure. After sorting, expert conducts manual analysis of the top N procedures, where N is selected in accordance with the available time of expert and the importance of detecting that the library is used. In order to determine similarity between the two binary procedures, values of the proposed software metrics are collected. The paper proposes a total of 19 different software metrics. The values of software metrics are compared with one out of the three proposed comparators, thus obtaining partial similarity measures...

Беланчевине или протеини су важни биолошки макромолекули полимерне природе (полипептиди), који се састоје од амино киселина и представљају основну градивну јединицу сваке ћелије..., Proteins are signicant biological macromolecules of polymeric nature (polypeptides), which contain amino acids and are basic structural units of each cell. Their contents include 20+3 amino acids and, as a consequence, they are presented in biological databases as sequences formed from 23 dierent characters. Proteins can be classied based on their primary structure, secondary structure, function etc. One of possible classications of proteins by their function is related to their contents in a certain cluster of ortholologous groups (COGs). This classication is based on the previous comparison of proteins by their similarities in their primary structures, which is most often a result of homology, i.e. their mutual (evolutionary) origin. COG database is obtained by comparison of the known and predicted proteins encoded in the completely sequenced prokaryotic (archaea and bacteria) genomes and their classication by orthology. The proteins are classied in 25 categories which can be ordered in three basic functional groups (the proteins responsible for: (1) information storage and processing; (2) cellular processes and signaling; and (3) metabolism), or in a group of poorly characterized proteins. Classication of proteins by their contents in certain COG category (euKaryote Orthologous Groups- KOG for eukaryotic organisms) is signicant for better understanding of biological processes and various pathological conditions in people and other organisms. The dissertation proposed the model for classication of proteins in COG categories based on amino acid n-grams (sequences of n- length). The set of data contains protein sequences of genomes from 8 dierent taxonomic classes [TKL97] of bacteria (Aquicales, Bacteroidia, Chlamydiales, Chlorobia, Chloroexia, Cytophagia, Deinococci, Prochlorales), which are known to have been classied by COG categories. The new method is presented, based on the generalized systems of Boolean equations, used for separation of n-grams characteristic

Proteini predstavljaju najvazniju grupu biomolekula u zivom svetu. Ra- zlicite funkcije koje imaju u svakom organizmu jedinstvene su i nezamenljive, pocev od raznovrsnih celijskih procesa, preko njihove strukturalne uloge, uloge katalizatora, velikog broja metabolickih funkcija i slicno. Poznavanje i razumevanje funkcija proteina je stoga esencijalno u istrazivanju bilo kog bioloskog procesa, sa posebnim naglaskom na oboljenja ljudi, s obzirom da se mnoga od njih mogu pojaviti zbog funkcionalnih mutacija. U ovom radu bice predstavljeno istrazivanje ovog domena kroz dva razlicita pristupa. U prvom, funkcija proteina posmatrana je kroz GO ontologije, koje podrazumevaju predstavljanje funkcije proteina kroz tri velika usme- rena aciklicka grafa funkcija: jedan je vezan za bioloske procese, drugi za celijske komponente, a treci za molekulsku funkciju. Svaki od njih sadrzi vise hiljada cvorova, pri cemu svaki cvor odreduje specicniju funkciju od svojih predaka. Zadatak ovog dela istrazivanja je razvoj prediktora funkcije proteina na osnovu njene primarne sekvence primenom metode struktural- nih podrzavajucih vektora koja predstavlja generalizaciju poznate metode podrzavajucih vektora na strukturalni izlaz. Jedno od osnovnih nacela molekularne biologije predstavlja paradigma struk- tura-funkcija po kojoj je 3D struktura proteina blisko povezana sa njegovom ulogom u organizmu. Utvr eno je da su neure eni proteini (kojima ne- dostaje 3D struktura) kao i neure eni delovi proteina u vezi sa teskim savre- menim bolestima i kao takvi su predmet aktuelnih istrazivanja. U drugom pravcu razmatrana je veza funkcionalnih kategorija proteina sa njihovom neuredenoscu, kao i sa drugim zicko-hemijskim karakteristikama proteina. Funkcija proteina ovde je posmatrana kroz 25 osnovnih funkcija proteina koje su grupisane u 4 funkcionalne grupe. U ovoj disertaciji su prikazani rezul- tati detaljne analize nad velikim skupom proteina za koje je neure enost odre ena primenom javno dostupnih ala, Proteins represent the most important groups of biomoleculs. Dierent functions that they carry out in each organism are unique and irreplaceable, including versatile cellular processes, structural role of proteins, catalytic function, a number of metabolic functions and so on. Knowing and under- standing protein function is therefore essential in investigation of any biolo- gical process, especially of human diseases since a lot of them are caused by functional mutations. In this paper, we represent investigation of protein function domain through two dierent approaches. In the rst one, protein function is represented by GO ontologies with the structure of a directed acyclic graph. There are three GO ontologies: one for functions regarding biological processes, one for functions regarding cellular components and one for molecular functions. Each ontology contains several thousands of nodes, where every node deter- mines more specic function than his ascendants. The task of this part of research was to develop a software for predicting protein function from its primary sequence based on structural support vector machines method which represents generalization of well-known support vector machines method on structural output. Structure-function paradigm is one of basic concepts in molecular biology, stating that 3D proten structure is closely connected to its role in organism. It has been detected that disordered proteins (the ones that lack 3D struc- ture) and disordered regions of proteins are related with severe contemporary illnesses, which contributed to their popularity in modern research. In an- other aspect, we investigated the relationship between proteins' functional categories and their disorder, as well ad with other physico-chemical char- acteristics of proteins. Here, protein function has been observed through 25 elementary functions grouped in 4 functional groups. In this work, we present results of thorough analysis over large protein dataset where dis

Intrinsically disordered proteins exist in highly flexible conformational states linked to different protein functions. In this work, we have presented evidence that HLA class-I- and class-II-binding T-cell epitopes, experimentally verified in several tumor-associated antigens and nuclear systemic autoantigens, are predominantly located in ordered protein regions or at disorder/order borderlines, defined by the majority of analyzed publicly available disorder predictors. We have also observed the overlapping of secondary structural elements and prevalently hydrophobic regions with T-cell epitopes in Epstein Barr Virus (EBV) nuclear antigen 1 (EBNA-1), cancer/testis antigen MAGE-A4, and Sm-B/B', U1 snRNPA (U1A) and U1-70 kDa autoantigens. The results are in accordance with the clustering of the predicted HLA class-I and class-II epitopes in protein parts which encompass the consensus of ordered regions, determined by individual disorder predictors. Some HLA class-II epitopes and linear B-cell epitopes were located near the segments predicted to have elevated crystallographic B factor in EBNA-1, Sm-B/B' and U1 snRNP A proteins, suggesting that protein flexibility could influence the structural availability of epitopes. Naturally processed T-cell epitopes and linear B-cell epitopes could also be found within putative disordered binding sites, determined by "dips" in the prevalently disordered parts of prediction profiles of the majority of disorder predictors, and peaks in ANCHOR-prediction profile. Two minor antigenic regions within EBNA-1, mapped to the residues 58-85 and 398-458, encompassing putative disordered binding sites, contain epitopes connected with anti-Ro 60 kDa and anti-Sm B/B' autoimmunity in systemic lupus erythematosus. One of these regions overlaps residues 395-450, identified as the binding site of USP7 (HAUSP), which regulates the EBNA-1 replication function. In Sm-B/B', one of the putative disordered binding sites (residues 114-165) encompasses th

Highly disordered protein regions are prevalently hydrophilic, extremely sensitive to proteolysis in vitro, and are expected to be under-represented as T-cell epitopes. The aim of this research was to find out whether disorder and hydropathy prediction methods could help in understanding epitope processing and presentation. According to the pan-specific T-cell epitope predictors NetMHCpan and NetMHCIIpan and 9 publicly available disorder predictors, frequency of epitopes presented by human leukocyte antigens (HLA) class-I or -II was found to be more than 2.5 times higher in ordered than in disordered protein regions (depending on the disorder predictor). Both HLA class-I and HLA class-II binding epitopes are prevalently hydrophilic in disordered and prevalently hydrophobic in ordered protein regions, whereas epitopes recognized by HIA class-II alleles are more hydrophobic than those recognized by HLA class-I. As regards both classes of HLA molecules, high-affinity binding epitopes display more hydrophobicity than low affinity-binding epitopes (in both ordered and disordered regions). Epitopes belonging to disordered protein regions were not predicted to have poor affinity to HIA class-II molecules, as expected from disorder intrinsic proteolytic instability. The relation of epitope hydrophobicity and order/disorder location was also valid if alleles were grouped according to the HIA class-I and HLA class-II supertypes, except for the class-I supertype A3 in which the main part of recognized epitopes was prevalently hydrophilic. Regarding specific supertypes, the affinity of epitopes belonging to ordered regions varies only slightly (depending on the disorder predictor) compared to the affinity of epitopes in corresponding disordered regions. The distribution of epitopes in ordered and disordered protein regions has revealed that the curves of order-epitope distribution were convex-like while the curves of disorder-epitope distribution were concave-like. The percenta

Упркос великом развоју вештачке интелигенције, људски мозак је још увек супериорнији узимајући у обзир разумевање и манипулацију делимично познатим чињеницама. Једна од области у којој је ово нарочито истакнуто односи се на проблем одговарања на питања. Када је потребно дати одговор, посебно онај који се експлицитно не налази у тексту корпуса који се претражује, предности експерта – човека представљају различите способности као што је способност објашњавања, комбиновања сложених одговора и апстрактног резоновања. Системи за Интелигентно Прослеђивање Питања (СИПП) имају за сврху размену знања на нивоу произвољне области експертизе и од значаја су за велики број апликација у којима се захтева интензивна комуникација између корисника. Корист од примене оваквих система укључује: (а) смањење непотребног оптерећења експерата који представљају вредан ресурс и (б) повећање квалитета услуга институције (универзитет, влада, предузеће), имајући у виду задовољство корисника с обзиром да су њихова питања прослеђена релевантним особама. У овом раду представљена је методологија пројектовања система за интелигентно прослеђивање питања написаних на природном језику – СИПП. На почетку је дат детаљан преглед ове области где је посебан акценат стављен на реализацију фаза СИПП процеса. Такође, у овом поглављу представљена је оригинална презентациона парадигма која генерализује суштину свих расположивих СИПП решења из отворене литературе. Презентациона парадигма садржи три основне фазе извршавања које се односе на три главна проблема приликом реализације система: анализу питања, прослеђивање питања и профилисање корисничког знања. На основу ове парадигме извршена је детаљна анализа и евалуација оваквих система, а као закључак наведен је предлог решења уочених проблема. У наставку дисертације описана је реализација предложених решења у виду прототипа СИПП система. У оквиру модула за обраду питања реализован је приступ који омогућава визуализацију питања, што обезбеђује интуитивну представ, In spite of great developments in artificial intelligence, human brain is still more powerful, concerning the comprehension and manipulation with partially known facts. One domain where this is prominent is related to a problem of question answering. When it comes to giving the answers, especially those that do not explicitly exist in the text corpus, the advantages of a human expert are abilities like explaining, combining complex answers, and abstract reasoning. Intelligent Question Routing Systems (IQRS) serve as a knowledge exchange medium in an arbitrary field of expertise, where intensive communication between users is required. The benefit coming from deployment of such systems includes: (a) reducing unnecessary “pinging” of experts, which are a valuable resource and (b) increasing the system owners’ (e.g. enterprise, government, university) quality of service, since users are more satisfied with answers, because their questions are answered by the right persons. This dissertation represents a methodology for IQRS systems design. It starts with a survey of the existing research in this domain, where the emphasis was put on the implementation of phases of the IQRS process. The survey also introduces an original presentation paradigm that generalizes the essence of approaches found in the open literature. The presentation paradigm includes three basic processing stages related to the three major problems of system implementation: question analysis, question forwarding, and users’ knowledge profiling. The outcome of this analysis is a proposal for a new approach that tackles identified problems. The rest of the dissertation describes an IQRS prototype which implements the proposed ideas. The question analysis module implements an approach that enables question visualization, thus it provides an intuitive representation of specific relations between concepts and their importance in the question. Also, this approach combines a fully automatic text processing and m

Frequently asked question in immunology is whether the immunodominance of T-cell epitopes is dependent to their localization in the antigen 3D structure or epitope-intrinsic. Using epitope prediction algorithms we have found that both HLA-I and HLA-II epitope frequencies were higher in ordered protein regions, for all analysed taxonomic categories (archaea, bacteria, eukarya, and viridae) and that epitopes appertaining to ordered protein regions were prevalently hydrophobic. Epitope frequency in disordered protein regions of various lengths was constant while in ordered regions had shown a rising trend with prolonging region length. The comparison between predicted and experimentally evaluated epitopes of several tumor associated antigens of cancer/testis antigen group, revealed that majority of epitopes presented by HLA-I and HLA-II molecules were localized in ordered protein regions.