Heatmap visualization design

heatmap-visualization-design

Summary

Generate color-encoded correlation heatmaps that depict relationships between metabolite signals and disease classes from deep learning model outputs. This visualization enables rapid identification of disease-associated metabolite signatures in mass spectrometry data.

When to use

After training a DeepMSProfiler model and generating per-sample predictions: when you need to display Pearson or Spearman correlation coefficients between individual metabolite signals and disease class labels in a matrix form suitable for publication or exploratory review of metabolite-disease associations.

When NOT to use

• Input is raw, un-normalized LC-MS spectra (not model-processed feature signals).
• Disease labels are continuous rather than categorical (use scatter plots or 2D density plots instead).
• Sample size is <10 per disease class (correlation estimates become unstable and heatmap becomes sparse).

Inputs

• Per-sample model outputs (sample predictions from trained DeepMSProfiler ensemble)
• Metabolite signal intensity matrix (features × samples)
• Disease class labels (categorical, one per sample)

Outputs

• Correlation heatmap image (PNG/SVG format)
• Correlation matrix (NumPy .npy array; shape: num_diseases × num_metabolites)
• Heatmap metadata (colorbar scale, correlation range)

How to apply

Load per-sample model outputs (metabolite signal intensities and disease predictions) from a trained DeepMSProfiler job directory. Compute correlation coefficients (Pearson or Spearman) between each metabolite feature and disease class labels across all samples. Construct a correlation matrix with metabolites as rows and diseases as columns, ensuring values are normalized for color mapping. Render the matrix using seaborn or matplotlib with a diverging colormap (e.g., coolwarm or RdBu) to encode both magnitude and direction of correlation; export as high-resolution PNG or SVG. The heatmap is typically generated as part of the feature extraction pipeline (run_feature) after model prediction completes.

Related tools

• seaborn (Python library for rendering correlation heatmaps with customizable color palettes and annotations)
• matplotlib (Underlying plotting backend for heatmap image generation and export)
• DeepMSProfiler (Deep learning framework that generates model outputs and feature extraction pipeline; heatmap generation is invoked via run_feature() or -run_feature flag) — https://github.com/yjdeng9/DeepMSProfiler

Examples

python showFeature.py

Evaluation signals

• Heatmap shape matches expected dimensions (num_metabolites × num_diseases); verify via .npy file shape inspection.
• Correlation values are bounded in [-1, 1] or [0, 1] depending on coefficient type; no NaN or inf values in matrix.
• Rows (metabolites) and columns (diseases) are labeled and legible; color intensity visually corresponds to magnitude (darker = stronger correlation).
• High-correlation metabolites cluster visually and align with known biomarkers or prior metabolomic literature for the disease context.
• Heatmap output file size and DPI are appropriate for publication (≥300 DPI for PNG; vector format preferred for SVG).

Limitations

• Correlation-based heatmaps assume linear relationships; nonlinear metabolite-disease associations may be masked.
• Pearson correlation is sensitive to outliers; Spearman is preferred for skewed metabolomics distributions but may reduce power with small sample sizes.
• Heatmap does not encode statistical significance (p-values); multiple-testing correction not applied by default in DeepMSProfiler.
• Feature ordering (row/column clustering) is not applied by default; manual reordering may be needed to reveal biological patterns.
• High-dimensional metabolite sets (>1000 features) produce unreadable heatmaps; filtering to top differentially correlated metabolites is recommended.

Evidence

• [other] Compute correlation coefficients (e.g., Pearson or Spearman) between each metabolite signal and disease class labels for all samples.: "Compute correlation coefficients (e.g., Pearson or Spearman) between each metabolite signal and disease class labels for all samples."
• [other] Construct a correlation matrix with metabolites as rows and diseases as columns.: "Construct a correlation matrix with metabolites as rows and diseases as columns."
• [other] Generate heatmap visualization using a Python plotting library (e.g., matplotlib, seaborn), with rows representing metabolites, columns representing diseases, and cell colors encoding correlation magnitude and direction.: "Generate heatmap visualization using a Python plotting library (e.g., matplotlib, seaborn), with rows representing metabolites, columns representing diseases, and cell colors encoding correlation"
• [readme] Heatmaps depicting the correlation of different metabolite signals with diseases.: "Heatmaps depicting the correlation of different metabolite signals with diseases."
• [readme] After run_feature ,the heatmaps were saved in ../jobs/jobs007/feature_results/ensemble_RISE.npy, so we can then show the feature heatmaps for different classes.: "After run_feature ,the heatmaps were saved in ../jobs/jobs007/feature_results/ensemble_RISE.npy"