Abstract 1777: Development of a pan-cancer 15 gene expression signature to detect a subgroup driven by MAPK signalling

Abstract

Abstract INTRODUCTION Unsupervised hierarchical clustering of gene expression data from 265 High Grade Serous Ovarian Cancer (HGSOC) patients identified 3 major molecular subgroups. One subgroup is driven by activation of the MAPK pathway and is associated with poor prognosis and resistance to platinum chemotherapy. The MAPK pathway is currently being targeted by novel therapeutics and hence an assay to detect activation of the pathway across cancers would be highly valuable as a clinical trial enrichment tool. Using internal and publicly available gene expression datasets we have demonstrated that the MAPK subgroup also exists in other cancer types and is associated with poor prognosis. The aim of this study was to develop a gene expression signature to predict the MAPK subgroup across multiple cancer types. METHODS Evaluation of gene expression data in a range of tumours (ovarian, colon, lung, melanoma and prostate) identified a common gene list (CGL) of 7802 genes showing high variability and high expression across diseases. The CGL was input to different machine learning algorithms developing signatures under 10x5-fold cross-validation (CV), trained against the MAPK HGSOC subgroup. Filter-Feature-Selection removed 10% of genes under CV based upon ranked correlation adjusted t-scores and the final model selected to satisfy a number of key criteria: AUC for predicting the endpoint; association with survival (C-Index); and functional relevance of signature content. RESULTS A 15 gene signature was selected, yielding an AUC=0.87 [95% CI:0.84-0.89] with respect to the MAPK subgroup. This model has validated as a poor prognostic marker in several other cancer types (Colorectal, Relapse free survival: HR=1.46 [95% CI:1.07-1.98]; Lung, Relapse free survival: HR=2.18 [95% CI:1.33-3.56]; Prostate cancer, Biochemical recurrence: HR 2.49 CI: 1.43-4.34), and is suppressed by MEK inhibition (p=0.0023) and elevated by KRAS, NRAS and MEK1 overexpression in cell line models (p=0.0443, <0.0001and <0.0001). Additionally we have demonstrated that the 15 gene signature strongly predicts response to the MEK inhibitors Trametinib and Selumetinib in established cell line models (p<0.001) and in primary cells isolated from breast and ovarian patients. CONCLUSION A 15 gene signature has been developed from formalin fixed paraffin embedded samples across multiple diseases to detect a molecular subgroup driven by MAPK signalling. This assay predicts sensitivity to MEK inhibitors in pre-clinical model systems and in primary cells derived from patients. Further work aims to validate the signature in clinical samples from patients treated with a MEK inhibitor. This assay may be helpful for clinical trial enrichment to select patients that are likely to benefit from MAPK targeted therapies. Citation Format: Laura A. Knight, Bethanie Price, Andrena McCavigan, Aya El-Helali, Charlie Gourley, Denis P. Harkin, Richard Kennedy, Nuala McCabe. Development of a pan-cancer 15 gene expression signature to detect a subgroup driven by MAPK signalling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1777. doi:10.1158/1538-7445.AM2017-1777