Journal: iScience

Article Title: Peripheral immune landscape in pancreatic ductal adenocarcinoma reveals expansion of effector states with disease progression

doi: 10.1016/j.isci.2026.115034

Figure Lengend Snippet: Dynamics of the peripheral immune population across PDAC progression Linear regression plots (left) and violin plots (right) illustrate the relationship between immune population abundance and PDAC progression. The selected immune populations include PD-1 + CD4 + T cells (A), central memory CD4 + T cells (B), early-like effector CD4 + T cells (C), CD45RA-CCR7- CD4 + T cells (D), T follicular helper cells (Tfh) (E), memory regulatory T (Treg) cells (F), and mature natural killer (NK) cells (G). In each linear regression plot (left), the x axis represents PDAC disease stages (1 = stage 1 PDAC patients, 2 = stage 2 PDAC patients, 3 = stage 3 PDAC patients, 4 = stage 4 PDAC patients), and the y axis represents the min-max normalized (MMN) percentages of a specific immune population. The blue regression line indicates the trend across disease stages. The coefficient of determination (R 2 ) and p value are displayed in red at the top of each plot. In each violin plot (right), the x axis compares healthy individuals (purple) and PDAC patients (red), and the y axis represents the original immune population percentage in total CD45 + cells or the absolute cell counts per mL of the whole blood. Each dot represents an individual sample, and the data are represented as mean ± SEM. The t test p value, indicating statistical significance, is displayed at the top of each violin plot. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001.

Article Snippet: PerCP Anti-Human CD45 (Clone: 2D1) , Cytek Biosciences , Cat# 67-9459-T100; RRID: AB_2621921.

Techniques:

Journal: iScience

Article Title: Peripheral immune landscape in pancreatic ductal adenocarcinoma reveals expansion of effector states with disease progression

doi: 10.1016/j.isci.2026.115034

Figure Lengend Snippet: Machine learning classification identifies predictive immune markers distinguishing PDAC from healthy individuals (A) Schematic overview of the machine learning workflow used to identify predictive immune markers from high-dimensional spectral flow cytometry data. Peripheral blood samples from 38 healthy individuals and 39 treatment-naive PDAC patients (stages I–IV) were analyzed. Mean fluorescence intensity (MFI) values of surface markers were extracted from CD45 + cells, asinh-normalized, and aggregated at the individual level. The dataset was split into 70% training and 30% testing sets for model development and evaluation. (B–D) Top 10 predictive markers ranked by feature importance scores across three independent classifiers: random forest (RF; B), gradient boosting (GB; C), and Bayesian additive regression tree (BART; D). Markers are ranked by their Gini importance scores. (E and F) Receiver operating characteristic (ROC) curves showing model performance in classifying healthy versus PDAC samples (E) and in predicting PDAC stages (F). Curves represent random forest (orange), GB (green), and BART (blue) classifiers, with area under the curve (AUC) values indicating model accuracy. (G and H) Overall expression levels of CD95 and CD45RA, and (H) the representative expression patterns of CD95 in PD1+ CD4 + T cells, CD45RA+ terminal effector (TE) CD8 + T cell, CD11c+ dendritic cells (DCs), and non-classical monocytes (Mo) of healthy individuals and PDAC patients. Each bar represents Z score normalized mean fluorescence intensity (MFI) across individuals. The data are represented as mean ± SEM, and the t test p value, indicating statistical significance, is displayed at the top of each violin and boxplot. ∗∗∗ p < 0.001; NS, non-significance.

Article Snippet: PerCP Anti-Human CD45 (Clone: 2D1) , Cytek Biosciences , Cat# 67-9459-T100; RRID: AB_2621921.

Techniques: Flow Cytometry, Fluorescence, Biomarker Discovery, Expressing