1219 Analysis of PD1+ T cells and lymphoid structure(TLS) presence in pleural mesothelioma using spatial and machine learning approaches

<h3>Background</h3> The relationship between tumor-infiltrating lymphocytes (TILs) and patient outcomes in pleural mesothelioma (MESO) is complex and not fully understood. PD1 expression and/or ligation can result in T cell activity modulation. Our study focuses on and lymphoid structures (LS) in the context of patient vitality, leveraging spatial analysis and machine learning techniques. <h3>Methods</h3> A cohort of 12 pathologist-confirmed epithelioid surgically resected tumor tissue from patients with pleural mesothelioma was analyzed following mIF staining, with attention given to CD3+PD1+ T cells and LS presence. LS presence was determined by using X and Y coordinates of T cells; through an algorithm (‘LS_finder’) developed in our lab and verified by pathologists. Patients with ambiguous or very low CD3 counts were excluded. Spatial analysis was conducted to evaluate the proximity and clustering of immune cells around tumor cells. Machine learning models, including logistic regression, Random Forest, SVM and XGBoost, were applied to assess the impact of various features on patient survival with a cutoff of May 2023. Bootstrapping was used to enhance the model performance. <h3>Results</h3> PD1+ T cell population results demonstrated a significant correlation with patient outcomes. The PD1+ T cells counts were higher (p=0.008) (Pearson r = 0.72) and the median (Pearson r = -0.3) and mean distances from tumor to PD1+ T cells were shorter in the survival group. The frequency of PD1+ T cells within 20microns from a tumor cell was lower in the survival group (p=0.0002). There was a modest correlation between LS presence and patient outcomes (Pearson r=0.47). Logistic regression and XGBoost classifiers identified key features influencing outcomes, with an optimized logistic regression model achieving a cross-validation score of 0.93, though it exhibited high standard deviation of 0.13 and a tendency to overfit. <h3>Conclusions</h3> The study highlights the relevance of PD1+ T cells in MESO patient outcomes and underscores the complexity of LS involvement. The integration of spatial analysis and machine learning offers a robust framework for understanding immune-tumor interactions, though further validation with larger cohorts is necessary. <h3>Acknowledgements</h3> This study was supported in part by the Dillon &amp; L Bourg Mesothelioma Fund, the Re &amp; RM Kennedy Lung Cancer Fund, the Fleming Endowed Fund, the MDA Rare Tumor Initiative, MDA Patient Mosaic, the NIH CCSG Award (CA016672 (Institutional Tissue Bank (ITB), Research Histology Core Lab (RHCL), and the TMP-IL MDA Moonshots Platform. We would like to thank the MDA Thoracic and Cardiovascular Surgery nurses and staff for their support of this study and the patients who consented to be included in the research protocol. <h3>Ethics Approval</h3> The study was performed under the MDA IRB approved LAB08-0380 protocol following patient informed consent. <h3>Consent</h3> The study was performed under the MDA IRB approved LAB08-0380 protocol following patient informed consent.