PL01.3.A Radiomic features and DNA methylation attributes in primary CNS lymphoma

Abstract Background Clinical and laboratory markers have been exploited to model risk in patients with primary CNS lymphoma (PCNSL), but the derived models do not fully explain observed variation outcome. Here we present an extended framework of phenotype-epigenotype correlations that reveal novel prognostic constellations enable prioritizing epigenetic therapy. Material Methods In this retrospective discovery validation study, leverage radiomic feature-driven analysis medical images supervised bioinformatic integration DNA methylation profiles. We integrate both data modalities synergistically using machine learning-based prediction cross-domain alignment. Ultimately, validate most relevant biological associations tumor tissues cell lines. Results a cohort 191 across 9 sites Austria external site South Korea, use T1-weighted contrast-enhanced magnetic resonance imaging derive score consists 20 mostly textural features. determine as strong independent predictive factor (multivariate HR=6.56), confirm its value cohort. Radiomic features align distinct, biologically meaningful ways, is predictable from selected (AUC=0.78). gene-regulatory differences between radiomically-defined groups converge on bcl6 binding activity, which posed testable treatment strategy subset patients. Conclusion The robust complementary predictor survival reflected at level PCNSL. Assessing selecting based phenotypes represents huge step forward, ability define provides concept advance modeling precision therapy for aggressive brain cancer.