Conversational Artificial Intelligence-Enabled Molecular Characterization of Sezary Syndrome Reveals Distinct Pathway-Level Alterations Compared with Non-Sezary Cutaneous T-Cell Lymphoma

BackgroundSezary syndrome (SS) represents an aggressive leukemic variant of cutaneous T-cell lymphoma (CTCL) with distinct clinical behavior compared with other CTCL subtypes. While prior studies have identified recurrent genomic alterations in CTCL, a systematic pathway-centric comparison between SS and non-SS CTCL remains limited. We applied our conversational artificial intelligence (AI) platform for precision oncology, to accelerate hypothesis generation and integrative interpretation of public genomic data. MethodsWe performed a secondary analysis of somatic mutation and clinical data from the Columbia University CTCL cohort available via cBioPortal. Samples were stratified into SS (n=26) and non-SS CTCL (n=17). High-impact coding variants were retained and annotated to curated functional gene groups and signaling pathways relevant to CTCL biology. Pathway-level mutation frequencies were compared using Fishers exact test, with effect sizes summarized by odds ratios. Tumor mutation burden (TMB) was compared using Wilcoxon rank-sum testing. Subtype-specific gene-gene co-mutation patterns were assessed using pairwise association testing and visualized with heatmaps and oncoplots, with our conversational AI agents facilitating interactive exploration and prioritization of results. ResultsOverall TMB did not differ between SS and non-SS CTCL (p=0.83), indicating comparable global mutational burden. Pathway-level analyses revealed enrichment of alterations affecting epigenetic regulators, tumor suppressor and cell-cycle control genes, NFAT signaling, and apoptosis/immune regulation in SS, whereas MAPK and JAK-STAT pathway alterations were relatively more frequent in non-SS CTCL. Co-mutation analysis demonstrated fewer but more focused gene-gene interactions in SS compared with broader co-mutation networks in non-SS CTCL, suggesting divergent evolutionary constraints. Several genes (including ERBB2, WWC1, POSTN) showed borderline subtype-specific enrichment, warranting further validation. ConclusionsConversational AI-enhanced analysis reveals that SS is distinguished from other CTCL subtypes not by higher mutational load, but by qualitative differences in pathway involvement, particularly epigenetic dysregulation, immune escape, and transcriptional control. These findings generate testable hypotheses for downstream validation in patient-level datasets and demonstrate the utility of conversational AI agents as accelerators of translational cancer genomics.