flow cytometry data analysis  (Tree Star Inc)

Journal: iScience

Article Title: Machine learning assisted immune profiling of COPD identifies a unique emphysema subtype independent of GOLD stage

doi: 10.1016/j.isci.2025.112966

Figure Lengend Snippet: Unbiased inflammatory profiling in combination with machine-learning reveals a highly divergent immune environment in COPD lungs with strong lymphocytic inflammation (A) Computational flow cytometry was performed on samples obtained from explanted lungs of patients with COPD ( n = 20) and healthy control samples ( n = 23) from downsized donor lungs (flow cytometry cohort), see for gating strategy. %CD45 + cells were taken for further analysis. (B) Stacked histogram showing relative global changes in immune cell distribution for dendritic cells (DC), macrophages, monocytes, lymphocytes, and polymorphonuclear leukocytes (PMNL) on a single patient level, see also A. (C) Principal component analysis (PCA) scores plot with biplot overlay representing the overall inflammatory profile consisting of 24 different cell populations from each lung and represented as one dot (grey-donors, red-COPD). (D) Supervised orthogonal projections to latent structures discriminant analysis (OPLS-DA) was directed toward the maximum difference between donors and COPD ( x axis) and intra-group differences on the y axis. Ellipses mark the 95% confidence interval of each group. (E) Representation of random forest (RF) analysis with 5,000 trees, model accuracy was evaluated with a split into 65% trainings set and 35% test set stratified for diagnosis. The contribution of each cell population to the RF model is illustrated by the distribution of its minimal depth (white boxes), lower value indicates higher importance. The color histograms represent the distribution how frequently and at what depth the cell type was used for splitting the trees. Cells are sorted in descending order of importance. For each population the log2 fold change (LFC) for each population is shown, dark red higher in COPD, gray higher in donor. (F) The multidimensional scaling (MDS) scores represent sample similarity and state the RF accuracy and 95% confidence interval. (G and H) The marked seven cell types occurring in >300 trees at root node were used for the simplified RF model and achieved similar accuracy. The distribution of these six cell types is shown in (H), Quantification via Wilcoxon rank-sum test with FDR multiple correction. %CD45 data was LOG-transformed as shown; ∗∗∗ p adj ≤ 0.001, black horizontal lines represent median values, see also C. (I) Representative immunofluorescence images of Donor and COPD formalin-fixed paraffin-embedded lung sections; nuclei = blue; T-cells = green, macrophages = yellow, B-cells = white, neutrophils = red, ∗indicates airways, see also . Scale bars represent 500 µm in overview panels and 100 µm in the zoom in sections. (J) Schematic summary of the changes in key immune populations. For each analyte, the direction of regulation is shown dark red higher in COPD, gray decreased in COPD.

Article Snippet: Flow Cytometry data , This paper , Mendeley Data: https://doi.org/10.17632/5f5k6dhgh5.1.

Techniques: Flow Cytometry, Control, Biomarker Discovery, Transformation Assay, Immunofluorescence, Formalin-fixed Paraffin-Embedded

Journal: iScience

Article Title: Machine learning assisted immune profiling of COPD identifies a unique emphysema subtype independent of GOLD stage

doi: 10.1016/j.isci.2025.112966

Figure Lengend Snippet: The cytokine networks in patients with COPD are distinct and compartmental specific (A) Cytokines were profiled in two separate cohorts (explorative and validation cohort) and the combined analysis shown. Peripheral blood ( n = 24 cytokines; samples: n = 30 control, n = 43 COPD) and lung homogenate ( n = 22 cytokines; samples: n = 36 donor, n = 42 COPD) samples from COPD and controls and subject to bioinformatical analysis, see also . (B–E) Principal component analysis (PCA) scores represent overall cytokine profile from lung homogenates and plasma colored (B) according to diagnosis (C), smoking history in controls (non—never smokers, ex—ex-smokers, and current smokers; D), and reported smoking history in pack years, sample with unknown history shown in gray (E). (F) Relative differences in compartmental cytokine levels between COPD and donors, higher values on x axis (lung) or y axis (plasma) indicates elevated in COPD. Mean values for each sample are shown and colored according to significance using Wilcoxon rank-sum test with FDR multiple correction. (G) Examples of cytokines differentially regulated between the lung and plasma. Plasma values are given as LOG-transformed concentration, and lung values as LOG-transformed concentration as standardized to protein concentration. Comparison by Wilcoxon rank-sum test ns p > 0.05, ∗∗ p ≤ 0.01, ∗∗∗ p ≤ 0.001 black horizontal lines represent median values, see also . (H) Schematic summary of compartmental cytokine changes, MVA, multivariate analysis; UVA, univariate analysis. The direction of regulation is shown for each analyte, dark red higher in COPD, gray decreased in COPD. (I) Multilevel correlation network constructed from pairwise correlations of significant circulating cytokines, lung cytokines, lung immune cells in the flow cytometry and explorative cytokine cohorts with clinical data. Correlations were calculated with Pearson correlation analysis using a cut off p ≤ 0.05 and | R |≥0.5 networks were visualized with Fruchterman-Reingold algorithm. Nodes represent individual parameters and edges were weighted by the corresponding correlation coefficients. Community detection was performed with a fast, greedy algorithm for the visualization of co-regulation patterns and the two detected communities are represented with gray shaded areas. The right community was primarily enriched circulating and clinical parameters, while the right consisted predominantly of lung parameters. (J) Visualization of selected correlations from (H), blue lines mark the Pearson correlation and gray ribbons the 95% confidence interval. Macs, macrophages; DC, dendritic cells; pO 2 , capillary partial pressure of oxygen; Smoking_py, smoking history in pack years.

Article Snippet: Flow Cytometry data , This paper , Mendeley Data: https://doi.org/10.17632/5f5k6dhgh5.1.

Techniques: Biomarker Discovery, Control, Clinical Proteomics, Transformation Assay, Concentration Assay, Protein Concentration, Comparison, Construct, Flow Cytometry

Journal: iScience

Article Title: Machine learning assisted immune profiling of COPD identifies a unique emphysema subtype independent of GOLD stage

doi: 10.1016/j.isci.2025.112966

Figure Lengend Snippet: Underlying immune signatures defines COPD sub-groups (A) Flow cytometry-based lung immune cell profiles in patients with COPD ( n = 20) were sub-grouped via K-means and Gaussian mixture models creating two subtypes A ( n = 12) and B ( n = 8). (B) Upper panel: Principal component analysis (PCA) scores representing the overall inflammatory profile (%CD45, LOG-transformed) consisting of 22 different cell populations from each lung and represented as one dot (grey-controls, red-COPD). Lower panel: The supervised method OPLS-DA was directed toward the maximum difference between COPD subtypes ( x axis) and intra-cluster differences on the y axis. Ellipses mark the 95% confidence interval of each group. (C) Separation of COPD subtypes based on the combined lung immune cell profiles (22 populations) and plasma cytokine (24 cytokines) levels peripheral blood (plasma) cytokines by PCA (upper panel) and OPLS-DA (lower panel). (D) Heatmap showing the relative changes for each analyte for each sample. Samples were hierarchically clustered, and presented as dendrograms. (E) Visualization of the effect size between the COPD sub-clusters for each cell population and circulating cytokine using Cohen’s-d, which standardizes the differences between two means and provides an estimate of the effect size, dot size reflects the −log 10 P adj value as determined by Wilcoxon rank-sum test with FDR multiple correction. (F and G) Top six regulated cell populations (F) and top five cytokine between COPD (G) subtypes as identified in D, ∗ p ≤ 0.05, ∗∗ p ≤ 0.01, ∗∗∗ p ≤ 0.001, as determined by Wilcoxon-Mann-Whitney-U-test, black horizontal lines represent median values. (H) Schematic summary of differences between the sub-types. For each analyte, the direction of regulation is shown for each analysis, dark red higher in subtype one.

Article Snippet: Flow Cytometry data , This paper , Mendeley Data: https://doi.org/10.17632/5f5k6dhgh5.1.

Techniques: Flow Cytometry, Transformation Assay, Clinical Proteomics, MANN-WHITNEY

Journal: bioRxiv

Article Title: A 3-genes interferon signature predicts sustained complete remission in pediatric AML patients

doi: 10.1101/2025.07.17.664572

Figure Lengend Snippet: Lymphocyte profiling of the L-TME in pediatric AML patients. (a) Logical gating strategy showing the T cell composition of L-TME in AML patients with (top row, n=4 ) and without (bottom row, n=5 ) sustained-CR. The data are displayed cumulatively in the dot plots and individually in the graphs (c and e). The same cumulative and individual representation is depicted in all figures showing flow-cytometry data. (b) I) Cumulative plots displaying CD45 + /CD34 − /CD19 − and CD3 − myeloid cells in AML patients with (top row, n=4 ) and without (bottom row, n=5 ) sustained-CR. II-III) Individual values of samples in (a) and (b) are displayed in a graph; Q value is the result of FDR-corrected multiple t-tests between sustained (red) and non-sustained (black) patients’ samples. Only significant differences are shown. IV) The % of blasts scored in the patients’ diagnosis for all the cohort is displayed in the graph for patients with and without sustained-CR. The blast score was calculated by a combination of microscope and clinical flow-cytometry analyses and provided as clinical report for patients in the Pathology Laboratory of Sidra. (c) I) Phenotype of the CD4 dim cell population: the majority of the cells are IFNγ + (shown in the colored Z-axis) CD4 + and CD8 + doublets, engaging in close networking interactions. A full panel of expression markers for this population is shown in Supplementary Figure 1b. II) Individual values of samples in (b) show a trend of enrichment in CD4 dim doublets (considering lower singlets) in patients without long-term remission.

Article Snippet: Statistical analyses of the flow cytometry data were performed using GraphPad Prism (v. 10.3.1).

Techniques: Flow Cytometry, Biomarker Discovery, Microscopy, Expressing