Soelistyo, Christopher Joseph;
(2023)
Learning biophysical determinants of cell fate in cell competition using interpretable deep neural networks.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Thesis_Christopher_Soelistyo_corrected.pdf - Accepted Version Download (73MB) | Preview |
Abstract
The development of machine learning in recent decades has raised the potential for a novel mode of scientific discovery. In this scheme, human scientists can investigate deep neural networks trained to model complex phenomena in order to gain insight into the base phenomena themselves. The work described in this thesis aims to assess this potential in the case of a specific test system: cell competition between wild-type and scribble-knockdown (scribkd) MDCK cells. Cell competition is a phenomenon in which less fit cells are removed from a tissue for optimal survival of the host. Prior research has uncovered certain aspects of cell competition in this system; for instance, that scribkd cells are eliminated via apoptosis triggered by mechanical compaction due to crowding. However, many questions remain, in particular, regarding the timing and precise combination of biophysical factors that influence these apoptotic events. In this thesis, I assess the potential of machine learning to gain insight into this phenomenon. I first demonstrate that deep neural networks can be trained to predict cell fate in scribkd before the fate event actually occurs, by using information about local cellular neighbourhood and morphology over time. I then explore several techniques to investigate the prediction-forming process of the trained networks. Strikingly, this approach recovers the result, demonstrated by several years of experimental research, that local cell density is the key driving factor behind cell competition in wild-type/scribkd MDCK cells. The work also reveals several challenges associated with this strategy, in particular, the issues of model interpretability and shortcut learning. I anticipate that the conclusions gleaned from this work may be of relevance to future attempts to leverage deep learning for scientific discovery.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Learning biophysical determinants of cell fate in cell competition using interpretable deep neural networks |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2022. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences > Div of Biosciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences > Div of Biosciences > Structural and Molecular Biology UCL |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10163664 |
Archive Staff Only
 |
View Item |
