What does this strategy enable?
This Strategy pairs registered microscopy with mass spectrometry imaging and fits a separate partial least squares relationship for each ion image. It goes beyond image overlays: optical features become predictors of molecular signals. The 2015 methodology formalizes this relationship for molecular tissue mapping, including lipid and small-molecule applications. Predicted distributions remain model outputs, not additional MS measurements or evidence of molecular identification.
Transferable workflow
Workflow
- 1
Acquire paired molecular and optical images
Measure MALDI ion images and microscopy of stained tissue from the same or an adjacent section. Preserve tissue coordinates and acquisition conditions.
- 2
Register images and derive paired features
Align tissue locations across modalities and derive microscopy descriptors such as hue, texture and entropy corresponding to the measured MS locations.
- 3
Fit ion-specific cross-modal models
Use partial least squares regression to model each ion intensity distribution from microscopy-derived predictors. Relationships and predictive performance differ between ions.
- 4
Evaluate reconstruction and spatial reliability
Examine reconstruction scores, absolute residual maps and bootstrap confidence-interval images for each ion. Do not use fusion to predict ions for which microscopy provides inadequate support.
- 5
Predict supported molecular distributions
Apply supported models to finer microscopy pixels or tissue regions with microscopy but no MS measurement. Label these outputs as predictions and retain the measured ion maps for comparison.
- 6
Interpret with measurement-based checks
Assess registration, sampling differences and model transfer before biological interpretation. Where feasible compare predictions with independent MS measurements; absence of optical support does not establish that an MS feature is noise.
Connected resources
Tools used in this strategy
Strategy profile
Components
Data acquisition, Preprocessing and feature extraction, Metadata analysis, Statistical analysis
Sample types
Tissue or biopsy
Biological context
Broadly applicable
Ion mobility
Ion mobility not required
Scope and considerations
Designed for
- Linking lipid or small-molecule tissue distributions with optical morphology.
- Predictive sharpening and spatial extrapolation when cross-modal relationships are supported.
Important requirements
- Spatially registered microscopy and MSI measurements with corresponding tissue structures.
- Ion-specific modeling and assessment of reconstruction, residuals and uncertainty.
Considerations
- Higher-resolution predictions do not increase the spatial resolution of the original MS measurements.
- Some ion distributions have no usable microscopy relationship and should not be predicted.
- Adjacent-section differences, registration errors and changed staining conditions can invalidate relationships.
- Extrapolation to unmeasured regions requires scrutiny of model transfer and overfitting.
- The original software is distributed through a licensing portal; open-source availability and execution have not been established.
Reusable resources
Publication
Van de Plas R, Yang J, Spraggins J, Caprioli RM. Nature Methods 12, 366-372 (2015).