Please use this identifier to cite or link to this item:

http://hdl.handle.net/10609/97527
Title: Comparison of clustering methods for multiparametric cytometry data analysis in order to implement an R/Shiny application
Author: Guadall Roldán, Anna
Director: Cánovas Izquierdo, Javier Luis  
Tutor: Adsuar Gómez, Antonio Jesús
Keywords: multiparametric cytometry
clustering
algorithm performance
Issue Date: 4-Jun-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: Conventional flow cytometry is an experimental technique enabling to measure up to 30 fluorescence parameters per cell. Recently, flow cytometry has been fused to mass spectrometry giving rise to a new methodology named mass cytometry that can potentially detect up to 100 parameters per cell. Cell populations are mainly characterized by a procedure known as gating, consisting in manually delimitating cell subsets using histograms or two-dimensional dot plots in a sequential manner. This procedure is time-consuming, imprecise and particularly inadequate to be used with a high number of parameters. In the past few years new computational techniques have been developed in order to efficiently handle high-dimensional cytometry data. However, such developments are still under evaluation. Furthermore, dealing with these techniques requires proficiency in using R packages and script writing. The main objective of this project is to provide cytometrists with efficient and easy-to-use unsupervised learning algorithms and visualization tools to explore high-dimensional cytometry data in a reproducible way. To that end, an extensive bibliographic research on unsupervised clustering algorithms applied to cytometry data has been performed and a methodology for performance evaluation has been developed. A selection of algorithms has been benchmarked using this methodology and both real cytometry and synthetic data, the latter being specially generated to that end. This comparative study has allowed the selection of a clustering algorithm, RPhenograph, to implement a Shiny application. The developed methodology is now ready to be applied to benchmark further algorithms and compare performances on other experimental designs.
Language: English
URI: http://hdl.handle.net/10609/97527
Appears in Collections:Bachelor thesis, research projects, etc.

Share:
Export:
Files in This Item:
File Description SizeFormat 
aguadallTFM0619memory.pdfMemory of TFM3.29 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons