Приказ основних података о дисертацији

Bioinformatics models for automatic prediction of human protein-protein interaction

dc.contributor.advisorPerović, Vladimir
dc.contributor.otherSavić-Pavićević, Dušanka
dc.contributor.otherVeljković, Nevena
dc.contributor.otherPerović, Vladimir
dc.contributor.otherSavić-Pavićević, Dušanka
dc.creatorŠumonja, Neven S.
dc.date.accessioned2020-02-07T10:29:23Z
dc.date.available2020-02-07T10:29:23Z
dc.date.available2020-07-03T08:08:05Z
dc.date.issued2019-11-01
dc.identifier.urihttps://nardus.mpn.gov.rs/handle/123456789/11826
dc.identifier.urihttp://eteze.bg.ac.rs/application/showtheses?thesesId=7180
dc.identifier.urihttps://fedorabg.bg.ac.rs/fedora/get/o:20874/bdef:Content/download
dc.identifier.urihttp://vbs.rs/scripts/cobiss?command=DISPLAY&base=70036&RID=51816975
dc.description.abstractInterakcije između bioloških makromolekula imaju ključnu ulogu u osnovnim procesima u živim organizmima, posreduju u metaboličkim putevima, putevima prenosa signala, transkripciji, translaciji i drugim ćelijskim i sistemskim procesima. Veliki broj oboljenja uzrokovan je mutacijama proteina u regionima odgovornim za interakciju sa drugim proteinima koje mogu dovesti do ometanja interakcije protein-DNK, promene u obrascima savijanja proteina, novih nepoželjnih interakcija i omogućavanje interakcije protein-patogen. Mapiranje interaktoma, odnosno kompletne mape interakcija protein-protein (IPP) unutar organizma, je od suštinske važnosti za razumevanje kompleksnih molekularnih odnosa unutar živih sistema, kao i za rasvetljavanje raznih patoloških stanja ljudskog organizma. Bioinformatičke metode za automatsko predviđanje IPP, kao suplementi eksperimentalnim metodama za analizu IPP, omogućavaju bolje razumevanje bioloških procesa i funkcija, lakše otkrivanje potencijalnih meta za ciljanu terapiju i smanjenja vremena i troškova razvoja novih terapeutika. U ovoj studiji razvijeni su modeli i metode za automatsko predviđanje IPP bazirane na mašinskom učenju i proteinskoj sekvenci, koja predstavlja univerzalnu, visoko kvalitetnu i eksperimentalno potvrđenu informaciju o proteinu. Generisani su modeli za predviđanje IPP za specijalne slučajeve: (i) između transkripcionih regulatora, odnosno proteina koji učestvuju u kompleksnom procesu transkripcione regulacije koji kontroliše ekspresiju gena i značajan je za normalnu fiziologiju ćelije, i (ii) proteina sa neuređenom tercijarnom strukturom, koji su kao takvi uključeni u ključne biološke procese interakcijom sa višestrukim partnerima, imaju fleksibilnu strukturu, višestruke funkcije, centralnu ulogu u regulaciji signalnih puteva, procesu prepoznavanja i vezivanja za male molekule, i čine većinu proteina povezanih sa neprenosivim bolestima. Pored toga, kreirane su tri nove vrste atributa za predstavljanje proteina: (i) atributi zasnovani na primarnoj strukturi proteina, (ii) evolutivni atributi i (iii) mrežni atributi, kao i metode bazirane na genetskom algoritmu za (i) automatsko generisanje i selekciju atributa i (ii) za automatsko formiranje i optimizaciju ansambla modela zasnovanim na mašinskom učenju, u svrhu proširenja prostora atributa i povećanja efikasnosti predviđanja IPP...sr
dc.description.abstractInteractions between biological macromolecules have a critical role in essential processes in living organisms, mediate the metabolic pathways, signaling pathways, transcription, translation and other cellular processes and systems. A large number of diseases caused by mutations in the regions of the protein responsible for the interactions with other proteins which can lead to interference of protein-DNA interaction, changes in the patterns of protein folding, new undesirable interaction and facilitate interaction of the protein-pathogen. Interactome mapping, i.e. mapping of complete network of protein-protein interactions (PPI) within the organism, is essential to an understanding of complex molecular relationships within the living system, as well as to elucidate the various human pathological conditions. Bioinformatics methods for automated PPI prediction, as addition to experimental methods for the analysis of PPI, allow a better understanding of biological processes and functions, easier detection of potential therapeutic targets, and reduce the time and cost of drug development. In this study, models have been developed and methods for the automated prediction of PPI based on machine learning and the protein sequence, which is a universal, high-quality and experimentally confirmed information on the protein. Models for the PPI prediction were generated for special cases: (i) between human transcriptional regulators, i.e. the proteins involved in the complex process of transcriptional regulation that controls the gene expression and they are important for normal cell physiology, and (ii) intrinsically disorder proteins, characterized by the lack of a fixed tertiary structure, which are as such involved in the regulation of key biological processes via binding to multiple protein partners, are malleable adapting to structurally different partners, have multiple functions, play a central roles in the regulation of signaling pathways, the process of molecular recognition and binding of small molecule, and are the prevailing protein class associated with noncommunicable diseases. In addition, three novel types of features for the representation of the proteins were created: (i) the features based on the protein sequence, (ii) the evolutionary features, and (iii) the graph features, as well as methods based on the genetic algorithm for (i) automatic feature-engineering process and (ii) automatic ensembling of different machine learning algorithms, in order to expand the feature space and to improve the PPI prediction performance...en
dc.formatapplication/pdf
dc.languagesr
dc.publisherУниверзитет у Београду, Биолошки факултетsr
dc.relationinfo:eu-repo/grantAgreement/MESTD/Basic Research (BR or ON)/173001/RS//
dc.rightsopenAccessen
dc.rights.urihttps://creativecommons.org/licenses/by-nd/4.0/
dc.sourceУниверзитет у Београдуsr
dc.subjectinterakcije protein-proteinsr
dc.subjectprotein–protein interactionsen
dc.subjectproteom čovekasr
dc.subjectproteinske sekvencesr
dc.subjectmašinsko učenjesr
dc.subjecthuman proteomeen
dc.subjectprotein sequenceen
dc.subjectmachine learningen
dc.titleBioinformatički modeli za automatsko mapiranje interakcija između proteina kod čovekasr
dc.title.alternativeBioinformatics models for automatic prediction of human protein-protein interactionen
dc.typedoctoralThesisen
dc.rights.licenseBY-ND
dc.identifier.fulltexthttps://nardus.mpn.gov.rs/bitstream/id/1792/Disertacija.pdf
dc.identifier.fulltexthttp://nardus.mpn.gov.rs/bitstream/id/1793/IzvestajKomisije21836.pdf
dc.identifier.fulltexthttps://nardus.mpn.gov.rs/bitstream/id/1793/IzvestajKomisije21836.pdf
dc.identifier.fulltexthttp://nardus.mpn.gov.rs/bitstream/id/1792/Disertacija.pdf
dc.identifier.rcubhttps://hdl.handle.net/21.15107/rcub_nardus_11826


Документи за докторску дисертацију

Thumbnail
Thumbnail

Ова дисертација се појављује у следећим колекцијама

Приказ основних података о дисертацији