Antonio Collesei

A subset of gastric cancers (GCs) is linked to Epstein-Barr virus (EBV) infection. This study aims to characterize the histopathological and molecular features of EBV-associated GCs (EBVaGCs), focusing on predictive biomarkers and genomic and transcriptomic analysis. A total of 35 primary EBVaGCs were considered. The presence of EBV was confirmed with in situ hybridization. Immunohistochemical analyses for HER2, PD-L1, claudin 18.2, and mismatch repair proteins were performed. Genomic and transcriptomic profiles were assessed using AmoyDx Master Panel, which can identify single-nucleotide variants, InDels, and copy number variations on 571 hot genes, as well as microsatellite status, tumor molecular burden, and homologous recombination deficiency at the DNA level; however, at the RNA level, it identifies rearrangements/fusions in 45 genes and also quantifies the expression of 2396 cancer-related transcripts. The following histotypes were identified: carcinoma with lymphoid stroma (CLS; 69%), tubular (20%), and mixed (11%). Most cases were associated with atrophic gastritis (71%), and only 11% with dysplasia. The vast majority (94%) of EBVaGCs expressed EBV-encoded RNA in all tumor cells. Mismatch repair deficiency and HER2 overexpression were each observed in 6% of cases, whereas all tumors had a PD-L1-combined positive score ≥10. Sixty-six percent of cases showed moderate/strong claudin 18.2 expression in ≥75% of cancer cells. The most frequently altered genes were PIK3CA (41%) and ARID1A (17%). Transcriptomic analysis revealed substantial differential gene expression between EBVaGCs and EBV-negative controls, with upregulation of genes involved in antigen presentation, natural killer cell-mediated cytotoxicity, and cytokine-cytokine receptor interaction in EBVaGCs. Within EBVaGC, CLS showed higher expression of immune-related transcripts and higher PD-L1 expression than other histotypes. This study establishes EBVaGC as a distinct molecular class, with a distinctive profile of genomic alterations and expression of predictive biomarkers, and also with a unique immune microenvironment with enhanced cytotoxic activity. The findings highlight EBV’s role in early tumor development and EBVaG-CLS as a distinct subgroup within EBVaGC, characterized by unique morphologic features and a pronounced immune activation profile.

Background HER2-low expression has gained clinical relevance in breast cancer (BC) due to the availability of anti-HER2 antibody–drug conjugates for patients with HER2-low metastatic BC. The well-reported instability of HER2-low status during disease evolution highlights the need to identify patients with HER2-0 primary BC who may develop a HER2-low phenotype at relapse. In response to the urgency of maximizing treatment access, we utilized artificial intelligence to predict this occurrence. Patients and methods We included a large multicentric retrospective cohort of patients with BC who underwent tissue resampling at relapse. The dataset was preprocessed to address relevant issues such as missing data, feature abundance, and target class imbalance. We then trained two models: one focused on explainability [Extreme Gradient Boosting (XGBoost)] and another aimed at performance (an ensemble of XGBoost and support vector machine). Results A total of 1200 patients were included in this study. Among 386 patients with HER2-0 primary BC and matched HER2 status at relapse, 42.5% (n = 157) converted to a HER2-low phenotype. The explainable model achieved a balanced accuracy of 58%, with a sensitivity of 53% and a specificity of 64%. The most important variables for this model were primary BC phenotype [mean Shapley value (SHAP) 0.540], primary BC histological type (SHAP 0.101), grade (SHAP 0.182), and sites of relapse (SHAP 0.008-0.213). The ensemble model had a balanced accuracy of 64%, with a sensitivity of 75% and a specificity of 53%. Conclusions This work represents one of the first proof-of-concept applications of machine learning models to predict a highly relevant phenomenon for drug access in modern BC oncology. Starting with an explainable model and subsequently integrating it with an ensemble approach enabled us to enhance performance while maintaining transparency, explainability, and intelligibility.