Metode predviđanja interakcija lijekova i ciljanih molekula

master's thesis

• Title (english): Methods for predicting drug-target molecule interactions
• Author: Petar Bratulić
• Mentor: Tomislav Šmuc (mentor)
• Committee member: Tomislav Šmuc (predsjednik povjerenstva)
• Committee member: Marko Horvat (član povjerenstva)
• Committee member: Nenad Antonić (član povjerenstva)
• Committee member: Zoran Vondraček (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Mathematics); Zagreb
• Defense date and country: 2021-03-04, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Mathematics

Abstract

U ovom radu analiziran je problem predviđanja interakcija lijekova i ciljanih molekula (DTI predviđanje). Predstavljeni su scenariji modeliranja interakcija lijekova i ciljanih molekula, kao i moderni algoritmi strojnog učenja koji se koriste u DTI predviđanju. Cilj rada bio je ocijeniti različite pristupe u kontekstu tih, različitih scenarija predviđanja interakcija. Algoritmi su evaluirani na različitim skupovima podataka, različitih veličina odnosno kompleksnosti. Za usporedbu algoritama, korištene su tipične metrike performansi koje se koriste u području strojnog učenja te je korištena i rigorozna statistička metodologija za usporedbu algoritama. Provedeni su eksperimenti u različitim scenarijima na različitim skupovima podataka čime smo došli do osnovnih zaključaka. Performanse svih metoda korištenih u radu uvelike ovise o scenariju problema. Promatrane metode daju najbolje rezultate u prvom najjednostavnijem scenariju, rekombiniranju lijekova. Prediktivna moć naučenih modela značajno je niža u scenarijima u kojem se model testira na novim lijekovima ili novim ciljanim molekulama.

Abstract (english)

In this thesis drug-target interaction problem has been analyzed. Scenarios for modelling drug target interaction problem have been introduced, as well as state-of-the-art machine learning algorithms. The goal of the thesis was to evaluate different approaches in context of said different interaction’s prediction scenarios. Algorithms were evaluated at benchmark datasets of different sizes and complexity. For comparison of algorithms, typical machine learning metrics were used as well as rigorous statistical methodology for algorithm comparison. By executing experiments in different scenarios on different datasets, we came to basic conclusions. Performances of all the methods used in the thesis depend largely on problem’s scenario. All observed methods give best results in the simplest scenario, drug repositioning. Predictive power of learned models is significantly lower in scenarios were new drugs or new targets were introduced.

• Keywords: DTI predviđanje; modeliranja interakcija lijekova; strojno učenje
• Keywords (english): drug-target interaction; machine learning algorithms
• Language: croatian
• URN:NBN: urn:nbn:hr:217:074538
• Study programme: Title: Computer Science and Mathematics; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra računarstva i matematike (magistar/magistra računarstva i matematike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2021-03-23 12:21:28