Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images
Thesis (MEng)--Stellenbosch University, 2021.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2021
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613990016778240 |
|---|---|
| access_status_str | Open Access |
| author | De Villiers, James Garrett |
| author2 | Theart, R. P. |
| author_browse | De Villiers, James Garrett Theart, R. P. |
| author_facet | Theart, R. P. De Villiers, James Garrett |
| author_sort | De Villiers, James Garrett |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2021. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/123778 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA |
| last_indexed | 2026-06-10T12:44:55.028Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2021 |
| publishDateRange | 2021 |
| publishDateSort | 2021 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/123778 Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images De Villiers, James Garrett Theart, R. P. Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. Mitochondrial -- Fission (Biology) Fusion and Depolarisation Fluorescence Microscopy UCTD Mitochondrial event localise Thesis (MEng)--Stellenbosch University, 2021. ENGLISH ABSTRACT: Cells are the basic building blocks of all living organisms and the overall health and quality of life are closely linked to their health and functioning. Amongst others, morphological mitochondrial events are quality control mechanisms involved in the maintenance of cellular health and viability. Mitochondrial fission, fusion and depolarisation events are highly coordinated events occurring simultaneously and continuously in cells. The inhibition of these events is characteristic of various age-related diseases such as Alzheimer’s disease and cancer. The ability to predict the occurrences and locations of these events provides insight into mitochondrial dynamics, which could in turn lead to the development of a means to intervene with one of the governing factors of the quality of life. Any method that can effectively predict or localise these mitochondrial events, will greatly enhance the research potential of mitochondrial dynamics. The mitochondrial event localiser (MEL) is a recently developed analytical method, which localises mitochondrial fission, fusion and depolarisation events in three dimensions. MEL was used to generate the ground-truth dataset that was used to train deep neural networks to predict the locations of these events. To reduce the processing time and eliminate the image pre-processing steps of MEL, the method was also recreated with deep neural networks in a two-dimensional domain. In this study, two methods were developed for the prediction of mitochondrial events in three dimensions, which uses a single frame of a time-lapse sequence as input. Another two methods were developed for the localisation of these events in two dimensions using two successive frames of a time-lapse sequence as input. The three-dimensional Pix2Pix generative adversarial network (GAN) and the Vox2Vox GAN were implemented for the prediction of the occurrences and exact locations of mitochondrial fission, fusion and depolarisation events in three dimensions using a single frame from a time-lapse sequence as input. Of the two methods, the Vox2Vox GAN yielded the highest location prediction accuracies. The Siamese Pix2Pix and SPADE GANs were implemented to localise mitochondrial fission, fusion and depolarisation events in two dimensions. The Siamese Pix2Pix GAN is a novel combination of the Siamese neural network and the Pix2Pix GAN, which receives two successive frames of a time-lapse sequence as input. The SPADE GAN receives difference images as input, which are constructed from two successive frames of a time-lapse sequence. Of the two methods, the Siamese Pix2Pix GAN yielded the highest localisation accuracies. In future, these methods can be modified for three-dimensional applications. This study confirmed that deep neural networks can localise and predict the occurrence and exact location of mitochondrial fission, fusion and depolarisation events. To our knowledge, this has never been done before and the results obtained can be viewed as state of the art. The neural networks developed for localisation and location prediction of mitochondrial events are high-throughput analysis methods to investigate mitochondrial dynamics. Further refinement of these neural networks can lead to the eventual implementation of these tools in the workflow of researchers in the field of life sciences. AFRIKAANSE OPSOMMING: Alle lewende organismes se algemene gesondheid en lewensgehalte is direk gekoppel aan die gesondheid en funksionering van die biologiese selle waaruit hulle bestaan. Morfologiese mitochondriale gebeurtenisse is van die kwaliteitsbeheermeganismes wat betrokke is by die handhawing van sellulêre gesondheid en lewensvatbaarheid. Mitochondriale splitsing, samesmelting- en depolarisasie-gebeurtenisse is hoogs gekoördineerde gebeurtenisse wat gelyktydig en voortdurend in selle voorkom. Daar is ’n sterk verband tussen die belemme- ring van hierdie gebeurtenisse en verskillende ouderdomsverwante siektes, soos Alzheimer se siekte en kanker. Die vermoë om hierdie gebeurtenisse en hul liggings te voorspel, bied ’n manier om insig en moontlike beheer oor van die bepalende faktore van lewensgehalte uit te oefen. Enige metode wat hierdie mitochondriale gebeurtenisse effektief kan voorspel of lokaliseer, sal dus die navorsingsmoontlikhede van mitochondriale dinamika aansienlik verbeter. Die mitochondriale gebeurtenislokaliseerder (MEL) is ’n analitiese metode wat mi- tochondriale splitsing-, samesmelting- en depolarisasie-gebeurtenisse in drie dimensies lokaliseer. MEL is gebruik om die grondwaarheidsdatastel te genereer wat gebruik is om diep neurale netwerke op te lei om die ligging van hierdie gebeurtenisse te voorspel. Om die verwerkingstyd te verminder en die voorverwerkingsstappe van MEL uit te skakel, is die metode ook herskep met diep neurale netwerke in ’n twee-dimensionele omgewing. In hierdie studie was twee metodes ontwikkel vir die voorspelling van die voorval en liggings van mitochondriale gebeurtenisse in drie dimensies, en nog twee metodes vir die lokalisering van hierdie gebeurtenisse in twee dimensies. Die drie-dimensionele Pix2Pix teenstrydig-skeppende netwerk (GAN) en die Vox2Vox GAN is geïmplementeer vir die voorspelling van die voorval en liggings van die mito-chondriale fisie-, samesmelting- en depolarisasie-gebeurtenisse. ’n Enkele raam van ’n tydsverloop-volgorde is as invoer gebruik vir hierdie netwerke. Van hierdie twee metodes, het die Vox2Vox GAN die hoogste akkuraathede vir die liggingvoorspellings gelewer. Die Siamese Pix2Pix GAN en SPADE GAN was geïmplementeer om mitochondriale splitsing-, samesmelting- en depolarisasie-gebeurtenisse in twee dimensies te lokaliseer. Die nuutgevonde Siamese Pix2Pix GAN is ’n kombinasie van die Siamese neurale netwerk en die Pix2Pix GAN, wat twee opeenvolgende rame van ’n tydsverloop-volgorde as invoerontvang. Die SPADE GAN ontvang verskil-prente as invoer, wat saamgestel is uit twee opeenvolgende rame van ’n tydsverloop-volgorde. Van die twee metodes het die Siamese Pix2Pix GAN die hoogste akkuraathede vir gebeurtenislokalisering gelewer. In die toekoms kan beide hierdie metodes aangepas word vir drie-dimensionele toepassings. Hierdie studie het bevestig dat diep neurale netwerke die voorval en ligging van mitochondriale splitsing-, samesmelting- en depolarisasie-gebeurtenisse kan voorspel en lokaliseer. Sover ons kennis strek is dit nog nooit vantevore gedoen nie en kan die metodes en resultate wat hier bespreek is as baanbrekerswerk beskou word. Die neurale netwerke wat gebruik word vir lokalisering en liggingvoorspelling van mitochondriale gebeurtenisse, is hoë volume deurvoermetodes vir die analise van mitochon- driale dinamika. Verdere verfyning van hierdie neurale netwerke kan lei tot die uiteindelike implementering van hierdie instrumente in die alledaagse bedryf van navorsingswerk in die lewenswetenskappe. Masters 2021-11-16T12:51:15Z 2021-12-22T14:20:54Z 2021-11-16T12:51:15Z 2021-12-22T14:20:54Z 2021-12 Thesis http://hdl.handle.net/10019.1/123778 en_ZA Stellenbosch University 160 pages application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Mitochondrial -- Fission (Biology) Fusion and Depolarisation Fluorescence Microscopy UCTD Mitochondrial event localise De Villiers, James Garrett Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title | Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title_full | Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title_fullStr | Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title_full_unstemmed | Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title_short | Mitochondrial fission, fusion and depolarisation event location prediction for a high-throughput analysis of three-dimensional fluorescence microscopy images |
| title_sort | mitochondrial fission fusion and depolarisation event location prediction for a high throughput analysis of three dimensional fluorescence microscopy images |
| topic | Mitochondrial -- Fission (Biology) Fusion and Depolarisation Fluorescence Microscopy UCTD Mitochondrial event localise |
| url | http://hdl.handle.net/10019.1/123778 |
| work_keys_str_mv | AT devilliersjamesgarrett mitochondrialfissionfusionanddepolarisationeventlocationpredictionforahighthroughputanalysisofthreedimensionalfluorescencemicroscopyimages |