Use this URL to cite or link to this record in EThOS:
Title: Image informatics approaches to advance cancer drug discovery
Author: Warchal, Scott J.
ISNI:       0000 0004 7969 1552
Awarding Body: University of Edinburgh
Current Institution: University of Edinburgh
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
High content image-based screening assays utilise cell based models to extract and quantify morphological phenotypes induced by small molecules. The rich datasets produced can be used to identify lead compounds in drug discovery efforts, infer compound mechanism of action, or aid biological understanding with the use of tool compounds. Here I present my work developing and applying high-content image based screens of small molecules across a panel of eight genetically and morphologically distinct breast cancer cell lines. I implemented machine learning models to predict compound mechanism of action from morphological data and assessed how well these models transfer to unseen cell lines, comparing the use of numeric morphological features extracted using computer vision techniques against more modern convolutional neural networks acting on raw image data. The application of cell line panels have been widely used in pharmacogenomics in order to compare the sensitivity between genetically distinct cell lines to drug treatments and identify molecular biomarkers that predict response. I applied dimensional reduction techniques and distance metrics to develop a measure of differential morphological response between cell lines to small molecule treatment, which controls for the inherent morphological differences between untreated cell lines. These methods were then applied to a screen of 13,000 lead-like small molecules across the eight cell lines to identify compounds which produced distinct phenotypic responses between cell lines. Putative hits from a subset of approved compounds were then validated in a three-dimensional tumour spheroid assay to determine the functional effect of these compounds in more complex models, as well as proteomics to determine the responsible pathways. Using data generated from the compound screen, I carried out work towards integrating knowledge of chemical structures with morphological data to infer mechanistic information of the unannotated compounds, and assess structure activity relationships from cell-based imaging data.
Supervisor: Carragher, Neil ; Sanguinetti, Guido Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: automated image analysis ; cellular fingerprints ; breast cancer ; drug discovery ; machine learning ; numeric morphological features ; cell-based imaging data