Primjena dubokog učenja za određivanje tipova stanica dobivenih scRNA sekvenciranjem

master's thesis

• Title (english): Deep-learning based scRNA-seq cell type identification
• Author: Luka Cavalli
• Mentor: Mirjana Domazet-Lošo (mentor)
• Committee member: Mirjana Domazet-Lošo (predsjednik povjerenstva)
• Committee member: Boris Milašinović (član povjerenstva)
• Committee member: Danijel Mlinarić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-04, Croatia

Abstract

Problem klasifikacije stanica uključuje identifikaciju i kategorizaciju različitih tipova stanica na temelju njihovih ekspresijskih profila gena dobivenih iz jedne stanice scRNA sekvenciranja. To je ključno za razumijevanje stanične heterogenosti u tkivima, proučavanje razvoja i funkcije organa te identificiranje promjena povezanih s bolestima, što može voditi do napretka u dijagnostici i terapiji. U ovom radu su za klasifikaciju korištene metode strojnog učenja (stroj potpornih vektora i... More slučajne šume) te metoda dubokog učenja gdje je kao model korišten transformer. Objašnjena je arhitektura transformera te su navedeni skupovi podataka na kojima su trenirani modeli. Algoritmi strojnog učenja ostvareni su pomoću biblioteka scikit-learn, a arhitektura transformera pomoću radnog okvira PyTorch u programskom jeziku Python. Performanse transformera i algoritama strojnog učenja su naposljetku uspoređene s rezultatima predtreniranog modela scBERT na dva skupa podataka. Najbolju točnost od isprobanih metoda su ostvarivali SVM i transformer, ali scBERT je na oba skupa podataka davao bolje rezultate. Less

Abstract (english)

The problem of cell classification involves identifying and categorizing different cell types based on their gene expression profiles obtained from scRNA sequencing. This is crucial for understanding cellular heterogeneity in tissues, studying the development and function of organs, and identifying disease-related changes, which can lead to advancements in diagnostics and therapy. In this paper, machine learning methods (support vector machines and random forests) and a deep learning... More method using a transformer model were used for classification. The architecture of the transformer is explained, and the datasets on which the models were trained are listed. The machine learning algorithms were implemented using the scikit-learn library, and the transformer architecture using the PyTorch framework in the Python programming language. The performance of the transformer and machine learning algorithms was finally compared with the results of the pretrained scBERT model on two datasets. The best accuracy among the tested methods was achieved by the SVM and transformer, but scBERT provided better results on both datasets.
Less

• Keywords: bioinformatika; duboko učenje; transformer; SVM; slučajne šume; scRNA sekvenciranje; Python; PyTorch
• Keywords (english): bioinformatics; deep learning; transformer; SVM; Random Forest; scRNA sequencing; Python; PyTorch
• Language: croatian
• URN:NBN: urn:nbn:hr:168:345612
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-10-12 20:16:30