Self-Organizing Maps and Wind-Rose Charts in the Visualization and Analysis of Flow Cytometry Data Open Access

Abstract

Using Flow Cytometry (FCM) technology, multi-faceted measurements can be
taken of many individual particles, and thus it is often used in tissue sample
analysis. As the resulting data set can have over ten dimensions and millions of
points, analysis can be complicated and visualizing the data requires
significantly reducing the number of dimensions or condensing the volume of
the data. In studies where FCM data has been taken from multiple patients at
multiple times, comparing these data sets complicates the analysis. I developed
a software package (IFC Soft) for analyzing and visualizing FCM data using
Self-Organizing Maps (SOMs) and Wind-Rose Charts (WRCs). I demonstrate
the use of SOMs and WRCs on FCM data taken from two different transplant
studies. In each study, blood samples were taken from post-op transplant
patients at regular time intervals and FCM data were produced from each
blood sample. The data were first pre-processed by medical researchers using
FlowJo software to remove cellular debris and miscellaneous cells and save the
general cell types for investigation. I used SOMs to visualize and manually
cluster these cells from the different blood samples and save the amount of
each cell type in each sample. I visualized these results using WRCs and SOMs
to look for common trends among different classes of patients (eg. responded
well/poorly to the transplant). The SOMs proved to be useful for selecting
clusters of cells, but were difficult to use directly for finding patterns between
patients because they displayed more information than could be easily
managed and because of the variability of the patients. WRCs made from the
cluster summaries were found to be more useful for finding patterns, but still
provided too much information to easily extract patterns. The SOMs of the
summary data, on the other hand, were easy to use in finding patterns among
patient groups. The WRCs and original SOM could then be examined to
confirm the pattern and see greater detail in each sample.

Table of Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Flow Cytometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Wind-Rose Charts (WRCs) . . . . . . . . . . . . . . . . . . . . 5
2.2 Self-organizing maps (SOMs) . . . . . . . . . . . . . . . . . . 6
2.2.1 The Incremental SOM Algorithm . . . . . . . . . . . . . . . 8
2.2.2 The Batch SOM Algorithm . . . . . . . . . . . . . . . . . . . 10
2.2.3 SOM Feature Maps: U-Matrix and Hit Histograms . . . . 11
2.3 Wind-Rose Charts used for summarizing FCM data . . . . . 12
2.4 Self-Organizing Maps used with Flow Cytometry . . . . . . 13

3 IFC Soft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1 About . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4 Using IFC Soft to Analyze FCM data . . . . . . . . . . . . . . . 18
4.1 Protective Immunity Project (PIP) . . . . . . . . . . . . . . . 18
4.2 Lung Transplant Data Set . . . . . . . . . . . . . . . . . . . . .19
4.3 Pre-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
4.4 SOM of All Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.5 Selecting Clusters of Cells . . . . . . . . . . . . . . . . . . . . 21
4.6 Using Wind-Rose Charts to Compare Cluster Summaries . 28
4.7 Displaying Summary Results on an SOM . . . . . . . . . . . .28

5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Math and Computer Science
Degree	MS
Submission	Master's Thesis
Language	English
Research Field	Computer Science
Keyword	IFC Soft FCM Flow Cytometry Wind-Rose Chart Self-Organizing Map SOM
Committee Chair / Thesis Advisor	Hertzberg, Vicki S, Emory University Xiong, Li, Emory University
Committee Members	Mehta, Aneesh K, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Self-Organizing Maps and Wind-Rose Charts in the Visualization and Analysis of Flow Cytometry Data ()	2018-08-28 13:21:21 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions