Poster Presentation 35th Lorne Cancer Conference 2023

Using the CUP Prediction Algorithm (CUPPA) to inform clinical curation of whole genome and transcriptome sequencing of cancers of unknown primary (#251)

Camilla Mitchell 1 , Richard Rebello 1 , Wendy Ip 1 , Tharani Sivakumaran 2 , Shohei Waller 2 , Clare Fedele 1 , Owen Prall 2 , Catherine Mitchell 2 , Aidan Flynn 1 , Atara Posner 1 2 , Shiva Balachander 2 , SUPER-NEXT study team 1 2 , Tran Pham 1 , Penelope Schofield 2 3 , Charles Shale 4 , Peter Priestley 4 , Oliver Hofmann 1 , Linda Mileshkin 2 , Sean Grimmond 1 , Joseph Vissers 1 , Richard Tothill 1
  1. University of Melbourne Centre for Cancer Research, The University of Melbourne, Melbourne, Victoria, Australia
  2. Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  3. Swinburne University of Technology, Hawthorn, Victoria, Australia
  4. Hartwig Medical Foundation, University of Amsterdam, Amsterdam, The Netherlands

Cancer of Unknown Primary (CUP) is a metastatic cancer with no identifiable primary cancer, despite standardised workups including imaging and histopathological evaluation. Metastatic disease is generally treated based the cancer’s tissue of origin (ToO).  CUP patients are generally treated with non-specific empirical chemotherapies, with limited survival benefit.

Understanding a CUP’s genomic profile may assist a more specific diagnosis, find actionable genomic targets, and facilitate access to tumour stream-specific and targeted treatments. A comprehensive whole genome approach may be warranted for mutational profiling, but insufficient data exists to support this. Therefore, SUPER-NEXT, supported by The Advanced Genomics Collaboration, uses whole genome and transcriptome sequencing (WGTS) to measure improved diagnostic and clinical benefit over contemporary gene panel-based sequencing.

In SUPER-NEXT, CUP prediction algorithm (CUPPA), a machine-learning tool developed by Hartwig Medical Foundation, and manual variant curation are employed. CUPPA weighs multiple features from WGTS to predict ToO. Multiple classifiers, including somatic mutation position, driver features, RNA expression and splicing profiling combine to provide a classification. When a high-likelihood match occurs, curation proceeds as though tumour type was known.

By combining CUPPA and variant curation with clinical review, we aim to optimise WGTS interpretation to resolve a cancer’s ToO. We identified three areas of WGTS interpretation which are routinely used to inform cancer diagnosis.

  1. Detection of pathognomonic mutations. Specific mutations and gene-fusion events have been linked to rare cancers or molecular-subtypes of common cancers. WGTS detection of these alterations, combined with expert multi-disciplinary review. can augment rare cancer diagnosis and resolve cases outside the scope of current prediction algorithms.
  2. Curation of recurrent combinations of mutations. Genomic landscapes of many tumour types have been characterised in large cohort studies (e.g. TCGA, ICGC, GENIE). CUP WGTS curation allows comparison of individual tumour genomic profiles with known cancer genomics landscapes, to provide potential diagnoses.
  3. Detection of patterns of damage due to environmental exposures, such as UV or tobacco. WGS can accurately determine mutational signatures of individual CUPs, providing clues to the ToO and potential therapeutic vulnerabilities.

Case studies illustrating these approaches will be presented, highlighting the utility of CUPPA-informed, WGTS curation in solving CUP diagnostic conundrums.