Abstract 1642: Spatial proteomics with hyperplexed fluorescence imaging predicts risk of colorectal cancer recurrence and infers recurrence-specific protein-protein networks

Abstract

Introduction: Colorectal cancer (CRC) is the third leading cause of cancer related deaths in United States with recurrence after resection-with-a-curative-intent being frequently implicated in these deaths. The basis for CRC recurrence is not completely understood, is multifactorial, and involves dysregulation of heterocellular signaling among tumor cells and their microenvironment. Based on hyperplexed immunofluorescence imaging and novel computational analyses, we have developed a recurrence-risk prediction method that samples these signaling networks within the epithelial and stromal domains of the tumor microenvironment and provides improved performance over current state-of-the-art recurrence-risk prediction assays. Data: In the retrospective study presented here, we used 52 hyperplexed immunofluorescence biomarkers associated with either canonical oncogenic pathways, immune response, or colon cancer per se to spatially profile tissue microarrays obtained from resected tissue samples from 432 chemo-naive CRC patients. Results: Using epithelial- and stromal-domain expression and co-expression diversity of the biomarkers, our preliminary results predicted the risk of CRC recurrence with a concordance index of 0.91. We also generated training and validation sets from the CRC patient cohort and demonstrated that the area under the curve (AUC) of the prediction receiver operating characteristic (ROC) was 0.90. We utilized stratified bootstrapping to show that the AUC was stable with a standard deviation of 0.02. Significantly, the penalized model selection used within our method allowed us to infer epithelial and stromal-domain protein networks specific to the risk-of-recurrence from the underlying signaling networks. Despite the limited sampling intrinsic to tissue microarrays we were able to capture immune cell infiltration and the differential modulation of these outcome specific protein-protein networks. Conclusions: Our CRC recurrence-risk prediction method exploits our spatial proteomics computational pathology platform involving hyperplexed immunofluorescence imaging. This study demonstrates the potential of this paradigm to not only accurately predict risk of CRC recurrence but also to reveal the underlying systems pathophysiology. Inferring outcome- and domain-specific CRC protein networks will enable biomarkers mechanistically linked to disease progression to be determined and their causality corroborated. In turn, this knowledge can be used to inform optimal therapeutic strategies for individual patients. Citation Format: Shikhar Uttam, Andrew M. Stern, Samantha A. Furman, Filippo Pullara, Fiona Ginty, D. Lansing Taylor, S. Chakra Chennubhotla. Spatial proteomics with hyperplexed fluorescence imaging predicts risk of colorectal cancer recurrence and infers recurrence-specific protein-protein networks [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1642.