Biomarker discovery in relapsed acute myeloid leukaemia and the impact of CD109

Acute myeloid leukaemia (AML) is a blood cancer which has an overall survival rate of around 30%. As a patient is treated for AML, about 50% will achieve complete remission, of those 50% will relapse within three years, often this leads to death due to a developed treatment resistance of the cancerous cells. To facilitate better clinical outcomes in AML a prognostic score was developed to predict which patients are at a higher risk of relapse and those that respond well to current treatment options. With the aim of developing a robust workflow of biomarker discovery, several bioinformatics approaches were implemented with the objective of identifying novel biomarkers in AML. Gene expression datasets of AML patients were used to trial a variety of workflows to discover a suitable relapse prognostic score. The workflows trialled involved both machine learning and statistical approaches of biomarker discovery. Overall a panel of 9 biomarkers were discovered and used to predict relapse in AML. The gene CD109 was identified in the process of biomarker discovery and was able to independently predict relapse in AML, where high expression of CD109 was associated with relapse in AML and had previously been reported in other cancers as associated with progression and treatment resistance. The CD109 protein is a co-receptor for TGFBR1, and facilitates its internalisation and degradation, thus disrupting the TGFβ1 signalling pathway. As the CD109 gene was highly associated with relapse, this study aimed to identify its functional relevance in AML cell lines using shRNA mediated gene silencing. The cells with a reduced expression of CD109 were treated with a combination of cytarabine and TGFβ1 to determine the response with altered gene expression. Some gene expression changes were observed depending on the treatment condition, indicating different pathways are activated in response to the treatment. The prognostic score created with this study has the potential to predict relapse before it happens, allowing patients to be monitored more closely and empowering clinicians to use alternative or targeted treatments. By using this prognostic score, the relapse rate in AML could be reduced, and patients who do relapse can be detected sooner. This score has the potential to improve survival rate in AML through identification of high-risk patients and increased monitoring. The workflows developed to discover biomarkers can be applied to a magnitude of settings allowing clinical tools to be developed and the identification of key genes that are indicative of a condition