Professor Dean Fennell and Professor Hongji Yang
MPMs harbour predominantly TSGs, which present a therapeutic challenge owing to the need to identify actionable and synthetic lethal co-dependencies. In contrast, oncogenic drivers constitute gain of function targets suitable for conventional small molecule inhibition. However, this class of cancer gene has not been comprehensively investigated in MPM.
We are leveraging multiregional exome sequencing of early stage MPMs in the MEDUSA study (Mesothelioma Evolution: Deciphering drUgable Somatic Alterations), to elucidate repeated evolutionary trajectories in MPM through the application of machine learning . We identified clonal somatic alterations involving TSGs that are associated with early de-repression of an oncogenic pathway and evolutionary bottleneck involving TEAD transcription (-22q/NF2, LATS2) . Beyond this, although several genomic regions involving clonal amplification implicate oncogenic gains of function eg. 8q24 (MYC), a comprehensive census of oncogenic driver networks and their drugable potential has not yet been explored in mesothelioma. Understanding the extent of clonal oncogenic drivers and their timing during mesothelioma evolution, may reveal novel targets for effective therapy.
The aim of this project is to curate a census of clonal gain-of-function somatic alterations in MPM, establish their associated evolutionary timing, phenotype, impact on prognosis and drug sensitivity.
1. Application of phylogenetic transfer learning  to uncover oncogenic evolutionary trajectories, timing large scale somatic events and involving oncogenic gain of function. We will aim to detect active oncogenic signalling pathways via paired gene set enrichment analysis of our MEDUSA transcriptome interrogation and validate pathway activation in matched MEDUSA derived cell lines. Machine learning based feature extraction will be used to explore the morphological, genomic, immune microenvironment, and clinical correlates of MPM subsets harbouring oncogenic drivers. All analysis will be validated in orthogonal TCGA data.
2. Exploration of novel drug-gene interactions by leveraging our drug sensitivity data derived from an exome sequenced MPM screening panel. We will employ a custom machine learning pipeline (comprising multiple algorithms based on transfer learning, random forests, LASSO/ridge regression , and visual neural network based classifiers). Pipeline processing times will be significantly reduced by execution on a supercomputing platform (currently used for an existing DAF-HY collaboration).
IMPLICATIONS OF THE PROJECT
The goal of this PhD project will be to expand the next generation of molecular targets suitable for clinical translation in precision medicine trials.
Yap et al, Nature Reviews Cancer 17
, 475, 2017
Zhang et al, Nature Communications 12
, 1751, 2021
Fennell D. Nature 572
, 314, 2019
Kolluri et al, eLIFE 7