Analiza molekulskih deskriptora u kvantitativnim odnosima strukture i topljivosti tvari

master's thesis

• Title (english): Analysis of molecular descriptors in quantitative structure-solubility relationships
• Author: Valnea Sindičić Đuretec
• Mentor: Šime Ukić (mentor)
• Committee member: Šime Ukić (predsjednik povjerenstva)
• Committee member: Mirjana Novak Stankov (član povjerenstva)
• Committee member: Marijana Kraljić Roković (član povjerenstva)
• Granter: University of Zagreb; Faculty of Chemical Engineering and Technology; Zagreb
• Defense date and country: 2019-09-16, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Chemistry

Abstract

Upotreba kvantitativnih odnosa strukture i svojstava tvari (QSPR) jedna je od glavnih ekonomskih alternativa u eksperimentalnom testiranju molekulskih svojstava, kao što je topljivost tvari u vodi, log S. U ovom radu, izrađeni su QSPR modeli primjenjujući algoritme strojnog učenja, stabala odluke i nasumičnih šuma, koristeći programski jezik Python. Modeli su rađeni na setu od 1311 molekula; koristili su se različiti molekulski deskriptori i molekulski otisci. Usporedbom modela pokazalo se kako algoritam nasumičnih šuma ima bolju moć predviđanja od algoritma stabla odluke. Također se pokazalo da modeli koji koriste molekulske deskriptore imaju bolju moć predviđanja od onih koji koriste molekulske otiske. Kao najbolji izabran je model nasumičnih šuma uz uporabu molekulskih deskriptora. Razvijeni model mogao bi poslužiti za predviđanje topljivosti molekula kandidata prilikom razvoja novih lijekova kako bi se smanjili troškovi eksperimentalnih ispitivanja.

Abstract (english)

The use of quantitative structure-property relationship (QSPR) is one of the major economic alternatives in experimental testing of molecular properties, such as the aqueous solubility of a substance, log S. In this master thesis, QSPR models were developed using machine learning algorithms, decision trees and random forests. The modeling was performed using Python programming language. The models were developed on a set of 1311 molecules; different molecular descriptors and molecular fingerprints were used. Model comparisons showed that the random forest algorithm had better prediction power than the decision tree one. Models using molecular descriptors indicated better prediction power than those using molecular fingerprints. The model of random forests using molecular descriptors was the one with the best characteristics. Once developed, the QSPR solubility model could be applied in developing process of new drugs in order to reduce the cost of experimental trials.

• Keywords: QSPR; molekulski deskriptori; molekulski otisci; topljivost
• Keywords (english): QSPR; molecular descriptors; molecular fingerprints; solubility
• Language: croatian
• URN:NBN: urn:nbn:hr:149:923359
• Project: Number: IP-2014-09-7992; Title: Modeliranje okolišnih aspekata napredne obrade voda za razgradnju prioritetnih onečišćivala; Title: Modeling of environmental aspects of advanced water treatment for fegradation of priority pollutants; Acronym: MEAoWT; Leader: Tomislav Bolanča; Jurisdiction: Croatia; Funder: HRZZ; Funding stream: IP
• Study programme: Title: Applied Chemistry - Graduate study; specializations in: Environmental chemistry and green technologies, Advanced materials and technologies, Applied organic chemistry; Course: Applied organic chemistry; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra primijenjene kemije (magistar/magistra primijenjene kemije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-02-27 15:07:02