Clustering Mass Spectrometry Data Using Variational Autoencoders
Loading...
Authors
Mainguy, Tyler
Date
Type
thesis
Language
eng
Keyword
Machine Learning , Deep Learning , Mass Spectrometry
Alternative Title
Abstract
Mass spectrometry is an analysis technique used to investigate the molecular profile of both biological and non-biological samples. Mass spectrometry imaging (MSI) allows for the spatial mapping of these molecules directly from a physical sample. Clinical applications of MSI are limited, in part because of the difficulty of analyzing the high-dimensional data that is produced. MSI images are commonly analyzed by reducing the dimensionality of the data and performing unsupervised analysis.
In this work, we analyzed MSI data by using a deep neural network composed of a variational autoencoder and a Gaussian mixture model. The model was trained to jointly optimize the latent space representation of the MSI data and the cluster
assignments within the latent space. We also used a path-based gradient-analysis method to determine the contribution of each analyzed molecule to its corresponding cluster. Our study demonstrated the robustness of the proposed method when compared to conventional MSI data-analysis methods. Our gradient-analysis method was able to discover patterns of molecules that had distribution throughout the samples which matched their corresponding cluster assignments.
Description
Citation
Publisher
License
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
CC0 1.0 Universal
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
CC0 1.0 Universal