Modelling the association between cell lines and tumour subtypes in large-scale genomics datasets
Year: 2023
Type: dissertation
Abstract: Cell lines derived from various cancer tissues allow efficient exploration of cancer biology and the efficacy of various cancer treatments without the hurdles associated with entire model organisms.However, cancer subtype information is severely lacking on cell lines.It is important for cell lines to be matched with a specific type of cancer since specific cancer subtypes have highly individual responses to treatments.Thus, correct matching of tissues and cell lines would allow data from cell line experiments to be better translated into treatment plans for specific subgroups of cancer patients, especially in terms of drug combination targeting.This study sought to achieve this matching via data driven subtype labelling of 226 cell lines representing primary lung, breast, colorectal and ovarian cancers.Firstly, by integrating transcriptomics data of tumour and normal tissues with cancer cell lines, followed by feature selection via Mutual Information based filtering and forward selection using Ridge, Support Vector Machine and Gaussian Naive Bayes classifiers trained on primary tissue gene expression data.The uncertainty of the predictions was analysed using via applicability domain and by scoring the one-vs-rest class probability distributions.The results showed that some cell lines were confidently assigned to subtypes, while others did not achieve a reliable matching.These results could be used to guide the use of cancer cell lines in future studies.
Author: Mikkonen, Juho
Institution: University of Eastern Finland
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
Funder: Itä-Suomen Yliopisto
OpenAlex ID: W7215554668
