RGWAS: A Novel Genome-Wide Association Study to Link Genomics to Radiomics for Pancreatic Cancer

Abstract

Listen

Translate article

<h3>Purpose/Objective(s)</h3> Pancreatic cancer is a critical global health problem with the highest mortality rate among all major cancers. The 5-year survival rate remained at ∼10% despite decades of clinical and research efforts. Clearly, new and synergistic approaches are needed to battle this ferocious disease. In this study, a novel radiomics-based genome-wide association study (RGWAS) was conducted in a unique patient population, linking genomics to radiomics in pancreatic cancer. <h3>Materials/Methods</h3> A unique pancreatic cancer rapid autopsy dataset was included for the study, where comprehensive tissue samples were collected from tumors and normal organs within hours of patient death. On 28 patients, whole exome sequencing (WES) was conducted to identify significant somatic mutations with VarScan. Diagnostic dynamic contrast CT of abdomen was used to manually segment the primary tumor and metastatic lesions, from which ∼1000 radiomic features were extracted. Using LASSO, radiomic features most significantly associated with clinical endpoints such as overall survival and with interesting genetic variants (i.e., somatic SNVs) were identified. <h3>Results</h3> Based on WES, somatic mutations of KRAS, TP53, and CDKN2A occur at high rates in both primary and metastatic tumors. Interestingly, some highly occurring somatic mutations in the primary tumor, such as BRCA1 and MYC, were not top ranked in metastases. Instead, somatic mutations of IS2, the insulin receptor substrate 2, occur at a high rate in metastases. RGWAS found several significant peaks (P-value<10⁻⁵) for certain radiomic features, uncovering interesting correlations between genetic mutations and imaging phenotypes. Some of these radiomic features were consistently associated with genetic variants between primary and metastatic tumors. <h3>Conclusion</h3> A unique patient dataset with rapid autopsy tumor and normal tissue samples was analyzed using a novel integrative RGWAS approach. RGWAS linked relevant radiomic features to interesting oncogenes and clinical phenotypes. These findings provide promising genetic and radiomic feature candidates for follow-up work, and shed light on the mechanism of pancreatic cancer progression and metastasis.