There are many factors which influence the behavior of a cell, in particular a cancer cell. Investigating these factors and their interaction provides an insight into the behaviour of a cancer, and in turn, it’s varying sub-types. To classify patients based on these sub-types to improve the efficiency of treatment, and therefore survival rates, previous studies have mostly looked at one of the “omes” of the multi-omics data i.e. the genome, proteome, or transcriptome, among others. This, however, only provides one dimensional insight into cancer cell behaviour and what influences it.
The aim of this project is to incorporate numerous “omes” of the multi-omics data into one model. This data is very complex and multi-dimensional, requiring a lot of computational resources. During this project, we apply and evaluate various Deep Learning and Natural Language Processing (NLP) approaches on this data. In addition to this, we investigate Explainable AI (XAI) tools, to make the model more generalisable and trustworthy.
– Become familiar with Deep Learning, NLP, and XAI approaches, especially their previous application on cancer data
– Evaluate existing cancer sub-type classification models
– Investigate the weaknesses and limits of transparency and trustworthiness of DL, NLP methods
– Implement and evaluate some of the DL, NLP and XAI approaches on multi-omics data