magistrsko delo
Anže Gregorc (Author), Tomaž Curk (Mentor)

Abstract

Interakcije protein-RNA sodelujejo v mnogih bioloških procesih. Modeliranje in preučevanje molekul proteinov in RNA nam tako lahko pomaga pri razumevanju medsebojnega delovanja proteinov in RNA. V magistrskem delu smo izdelali postopek napovedovanja interakcij protein-RNA na proteinu z uporabo konvolucijskih nevronskih mrež nad grafi. Podatke smo pridobili iz podatkovne baze PDB, jih predelali v strukturo grafa in vsakemu atomu dodali primerne značilke. Tako so podatki primerni za modele nevronskih mrež, ki delujejo nad grafi. Modele smo ovrednotili in prikazali rezultate z različnimi merami uspešnosti. Najboljši model dosega dobre rezultate (ROC AUC = 0,9). Implementirali smo tudi grafični vmesnik, ki v 3D prostoru prikaže strukturo proteinov in napovedana mesta interakcij z RNA.

Keywords

molekulske interakcije;globoke nevronske mreže;konvolucijske nevronske mreže nad grafi;računalništvo in informatika;magisteriji;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [A. Gregorc]
UDC: 004.414.23:543.384(043.2)
COBISS: 83603715 Link will open in a new window
Views: 172
Downloads: 49
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Modeling protein-RNA interactions with deep graph convolutional neural networks
Secondary abstract: Protein-RNA interactions are involved in many biological processes. Modeling and studying protein and RNA molecules can help us understand the workings of proteins and RNA. In this master's thesis, we developed a procedure for predicting protein-RNA interactions on a protein using convolutional neural networks over graphs. We obtained the data from the PDB database, preprocessed it into a graph structure, and added appropriate features to each atom. Thus, the data are suitable for graph neural network models. We analyzed the models and presented the results with different performance metrics. The best model achieved good results (ROC AUC = 0.9). We also implemented a graphical interface to visualize the structure of proteins and the predicted sites of interaction with RNA in 3D space.
Secondary keywords: molecular interactions;machine learning;deep neural networks;graph convolutional neural networks;computer science;computer and information science;master's degree;Modeliranje podatkov (računalništvo);Strojno učenje;Beljakovine;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Master's thesis/paper
Study programme: 1000471
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 76 str.
ID: 13715149