EpiFlow D3

Multiparametric histone PTM profiling by spectral flow cytometry

Upload your .rds file to begin analysis

Ridge plots
Violin plots
LMM statistics
Heatmaps
Cell Cycle
Correlation
Positivity
Gating
PCA
UMAP + FeaturePlot
Clustering
ML

Data Overview

Cells per Condition

Cells per Replicate

Cells per Identity (by condition)

Cells per Cell Cycle Phase (by condition)

Cells per Replicate (by condition)

Condition × Cell Cycle Distribution

Marker Distribution Overview (median = dashed line, box = mean ± SD)

Marker Distribution by Condition (median = dashed line, box = mean ± SD)

Condition × Identity Cross-tabulation

Ridge Plot

Ridge plots show full H3-PTM intensity distributions. Tip: Group by Identity, color by Genotype to reveal population-specific epigenetic shifts — overlapping curves show direct WT vs mutant comparisons within each cell population.

Marker Expression

Violin plots show distribution shape with box plot summaries. Tip: Group by Identity, color by Genotype for side-by-side paired comparisons within each cell population.

Statistical Analysis

LMM accounts for nested data structure (cells within replicates). Replicate = random effect to avoid pseudoreplication. Stratify by Identity/Cell Cycle to find population-specific effects.

Identity Helper Heatmap

Z-score heatmap of marker expression per group. Blue = below average, Red = above average. Toggle "Include phenotypic" to add non-H3 markers. Use to identify cluster identities from marker signatures.

Cell Cycle Analysis

Cell cycle is a critical confounder: H3K27ac rises in S/G2 due to histone deposition. Tests whether cell cycle composition differs between conditions. Click a phase below to see per-marker H3-PTM comparisons within that phase.

Per-Phase Marker Analysis

Correlation Analysis

Global computes one correlation matrix across all cells. Per-Group + Differential computes separate matrices per genotype, then applies Fisher z-transform to test whether marker-marker correlations differ significantly between groups. By default, the Fisher z uses replicate-level N for proper inference. Check "Use cells as replicates" only if your experiment lacks biological replicates. Disrupted co-regulation (Δr with p<0.05) indicates chromatin regulator loss affects the coordinated deposition of marks.

Positivity / GMM

Fits a 2-component Gaussian Mixture Model (GMM) to identify positive (high) vs negative (low) subpopulations for each H3-PTM mark. The threshold is set at the GMM crossover point. Compares fraction-positive between genotypes using Fisher's exact test, and assesses distribution shifts via KS test. Leave "Manual threshold" blank for automatic GMM-based detection. This is key for marks with bimodal distributions (e.g. H3K27me3).

Quadrant Gating

vs
Interactive bi-axial scatter plot with draggable quadrant thresholds (blue crosshairs). Click a quadrant to see H3-PTM density profiles and cell cycle distribution for cells in that quadrant. Chi-square tests whether quadrant distributions differ between groups.

PCA / Batch Effect QC

PCA reduces multi-dimensional H3-PTM space. Color by Replicate to check for batch effects — replicates of the same genotype should cluster together.

UMAP

n_neighbors: low = local structure (tight clusters), high = global topology. min_dist: low = dense clusters, high = spread out. ℹ️ Hover for details.
UMAP preserves local neighborhoods to reveal cell population structure. Select a marker from the color dropdown to see fluorescence intensity (FeaturePlot). Toggle Split to compare genotypes side-by-side on the same embedding. Re-coloring uses cached data — no need to re-run.

Clustering

Clusters cells by H3-PTM profiles on a UMAP embedding. K-Means: specify k. Hierarchical: Ward linkage. Louvain/Leiden: graph-based, resolution controls granularity. Use Elbow Scan to find optimal k. After clustering, rename clusters below to assign biological identities, then Apply to update all tabs.
Compare with:

Volcano Plot

Run "All Markers" in Statistics tab first

Forest Plot

Each color = H3-PTM marker (to track modifications across subsets). Gray = not significant (p≥0.05). Red p-values = significant. Use "Generate" to run All Markers LMM with chosen stratification directly.

Machine Learning

RF and GBM classify cells using H3-PTM + phenotypic features. Uncheck noisy features (FxCycle, Live/Dead, etc.) below. Use "Run All" to compare all models side-by-side.

Random Forest

Click "Random Forest" to run

GBM (xgboost)

Click "GBM" to run

H3-PTM Signatures

Click "Signatures" to run

Diagnostic Signature Assessment

Assess whether H3-PTM signatures can discriminate genotype: MANOVA tests if multivariate profiles differ; LDA 5-fold CV estimates diagnostic accuracy; Consistency tests if effects hold within each identity (via LMM). Stratify to test cell-type or cell-cycle specificity.