Transcriptomic subtyping of prostate cancer

Firstly, using a data-driven approach, multiple primary prostate cancer datasets, comprising both gene expression and clinico-pathological data, were analysed using the Clustering Intra and Inter DatasEts (CoINcIDE) meta-clustering framework and an adaptation of the prior knowledge transfer tool Gene Expression Compositional (GECA). Subtypes were analysed in the context of survival. Using a knowledge-driven, dimension reduction approach with multiple primary prostate cancer datasets, six sets of biologically relevant signature scores, estimated using the claraT tool, were used as inputs for consensus clustering.

Clusters were evaluated, considering survival, biological characteristics, including differential gene expression, dysregulated pathways, and genomic alterations.<br><br>The clustering outputs of the CoINcIDE and GECA approaches failed to demonstrate consistent prognostic relevance. However, in the claraT approach, two subtypes were discovered: the Genome-Unstable subtype defined by relatively higher expression of the Genome-Instability signature scores and the Genome-Stable subtype associated with lower expression of the Genome-Instability signature scores.

The Genome-Unstable subtype also had a higher risk of metastatic recurrence, biochemical recurrence and iv was associated with higher Gleason scores. Of the claraT signature set, the published CIN25 signature was identified as the most accurate diagnostic biomarker of the subtypes. Mitotic dysregulation was identified as a mechanism potentially driving the genomic instability phenotype of the Genome-Unstable subtype.

Furthermore, the Genome-Unstable subtype was identified as potentially targetable by PARP inhibitors and/or POLQ inhibition due to dysregulation of DNA damage repair mechanisms and the upregulation of POLQ in this subtype.<br><br><i>Thesis is embargoed until 31 December 2027</i>.