'To apply, please click institution web link'*
Pancreatic cancer (PC) continues to be one of the deadliest cancers, predicted to soon become the 2nd cause of cancer death in our society. Most patients die of their disease within 1 year of diagnosis, making PC a disease of considerable unmet need. Therapeutic development has been challenging. This, in part, can be attributed to the rapid clinical deterioration of many PC patients, where the majority are unable to receive either standard of care combination chemotherapy, or treatment within clinical trials. The dynamics driving this rapid progression are complex, involving tumour-host interactions, high genetic variability, and the capability of cancer cells to adopt different identities along a phenotypic spectrum (lineage plasticity). Subtyping these different phenotypes based on clinically and biologically relevant molecular similarities is essential to better define molecular taxonomy of prognosis and treatment response. Clinically, haematoxylin and eosin (H&E) histopathological images continue to be the main modality for cancer diagnosis, their prognostication and potential treatment selection. These whole slide image (WSI) contains abundant information on how the tumour grows and their interactions with the tumour microenvironment, most of which are not routinely reported by pathologists. Recently, the field of histological deep learning applications has seen great progress in their ability to predict tumour classification, molecular aberration, survival, and immunotherapy response, not possible by human prediction. We hypothesize that we can use novel deep learning approaches to identify the histomorphological features on WSIs that drive the rapid progression phenotype of PC. This work will leverage the unique availability of data generated as part of Precision-Panc (https://www.precisionpanc.org/), with well-annotated clinicopathological information, molecular data and survival outcomes of patients who registered for the study but showed rapid decline and never started treatment or died within 8 weeks of starting chemotherapy (~40-50% of patients). The overall aim is to identify molecular mechanisms of rapid disease progression and define the histomorphologic phenotype clusters (HPCs) driving this rapid progression phenotype. The specific aims are to: 1. Define molecular driver events of rapid pancreatic cancer progression, using the prospectively accrued Precision-Panc cohort with well-annotated clinicopathological, bulk tumour genome, transcriptome and survival data available. 2. Identify histomorphologic phenotype clusters (HPCs) that drive the rapid progression phenotype, using deep learning approaches, and validate the HPCs and molecular processes by transcriptome sequencing of laser-capture microdissected areas of interest and multiplex immunofluorescence (mIF). 3.
Validate the histomorphological biomarker identified in Aim 2 in independent patient cohorts using the National Safe Haven database and other patient cohorts through the international collaboration of the supervisory team including PANCAIM EU patient cohorts (https://pancaim.eu) and International
Cancer Genome Consortium. Overall, this will result in a better understanding of PC progression with identification of potential therapeutic targets and a histomorphological candidate biomarker that can be used in routine clinical practice to guide patient care.
*You must download and complete the below form with your application.