eprintid: 31534
rev_number: 16
eprint_status: archive
userid: 6646
dir: disk0/00/03/15/34
datestamp: 2022-04-28 13:25:50
lastmod: 2022-05-02 11:53:36
status_changed: 2022-04-28 13:25:50
type: doctoralThesis
metadata_visibility: show
creators_name: Meechan, Kimberly
title: Tools for multimodal atlases: bridging morphology and gene expression
subjects: ddc-500
subjects: ddc-570
divisions: i-140001
divisions: i-850800
adv_faculty: af-14
abstract: Maps combine different types of data, from a number of different sources, into the uniform grid of their coordinate system. They are invaluable resources as they allow the connections between data to be understood in their spatial context. We have been making biological maps for decades of all kinds of systems - for example, model organisms like Drosophila melanogaster and Caenorhabditis elegans have large atlases of neuronal cell types available.

These large multimodal atlases (i.e. datasets that combine multiple data sources into one coordinate system) tend to undergo a similar ‘lifecycle’. First, a standard coordinate system is formed, followed by collection of the required datasets. These are then integrated into the common coordinate system, before they are analysed, and finally shared as a resource to the biological community.

Electron microscopy (EM) offers many advantages when integrated into atlases of this kind. Firstly, it can achieve very high resolution, making it ideal for visualising small features like the fine structure of different organelles. It also provides the full spatial context of a tissue - allowing the locations and orientations of cells to be understood in a comprehensive manner. This comes at the cost of a number of challenges though. For one, EM is quite low throughput, meaning that often only a single sample can be imaged, or only small regions of a sample. Also, EM data can be extremely large (many terabytes in size) making it difficult to store, analyse, and share with other researchers.

In my thesis, I aim to ease the integration of large-scale volume EM into multimodal atlases at various points within this lifecycle. First, at the acquisition stage, I look at increasing the speed, ease and precision of targeted EM acquisition. I develop two different methods - one focused on trimming resin-embedded blocks with an ultramicrotome, and the other with a Serial Block Face Scanning Electron Microscope (SBEM). Both these methods are provided as user-friendly Fiji plugins, making them accessible to those with no programming experience. 

Second, at the analysis stage, I create an analysis pipeline for comparing morphology (from EM) with gene expression of cell types in a large multimodal atlas. This focuses on a large atlas of the organism Platynereis dumerilii that combines volume EM with gene expression patterns for over 200 genes. This analysis provides full body clustering of cells based on morphology and gene expression, and shows a clear correspondence between gene expression and morphological tissue boundaries from the EM. 

Finally, at the resource sharing stage, I look at improving the ease of sharing and exploration of these massive EM datasets. To do this, I contribute to the software MoBIE that allows interactive browsing of very large, remotely stored image data. I focus on making the creation of MoBIE projects accessible to those with no programming experience, adding a user interface for creating projects via the Fiji plugin.
date: 2022
id_scheme: DOI
id_number: 10.11588/heidok.00031534
ppn_swb: 1800616406
own_urn: urn:nbn:de:bsz:16-heidok-315344
date_accepted: 2022-03-28
advisor: HASH(0x55fc36d7f7a0)
language: eng
bibsort: MEECHANKIMTOOLSFORMU2022
full_text_status: public
place_of_pub: Heidelberg
citation:   Meechan, Kimberly  (2022) Tools for multimodal atlases: bridging morphology and gene expression.  [Dissertation]     
document_url: https://archiv.ub.uni-heidelberg.de/volltextserver/31534/1/KimberlyMeechan_thesis.pdf