Patient Similarity Clustering - Discover Patient Subgroups

Visualizes patient similarity using dimensionality reduction techniques (PCA, t-SNE, UMAP, MDS). Projects high-dimensional patient data into 2D or 3D space to reveal natural patient groupings. Inspired by Orange Data Mining's interactive projection widgets, adapted for jamovi with comprehensive cluster analysis and statistical validation.

Details

This analysis performs dimensionality reduction to visualize patient similarity:

• PCA: Linear method preserving global variance structure

• t-SNE: Non-linear method excellent for visualization, preserves local neighborhoods

• UMAP: Non-linear method preserving both local and global structure, faster than t-SNE

• MDS: Classical scaling method preserving pairwise distances

Optional cluster analysis identifies patient subgroups using k-means, hierarchical clustering, or DBSCAN. Survival analysis can compare outcomes across discovered clusters.

Use Cases

• Discover unexpected patient subtypes based on clinicopathological features

• Identify which variables drive patient groupings

• Validate if known outcomes correspond to natural patient clusters

• Find prognostic patient subgroups for stratified treatment

References

Orange Data Mining: https://orangedatamining.com/widget-catalog/unsupervised/

Author

ClinicoPath Development Team

Super classes

jmvcore::Analysis -> ClinicoPath::patientsimilarityBase -> patientsimilarityClass

Methods

Public methods

• patientsimilarityClass$clone()

Inherited methods

• jmvcore::Analysis$.createImage()
• jmvcore::Analysis$.createImages()
• jmvcore::Analysis$.createPlotObject()
• jmvcore::Analysis$.load()
• jmvcore::Analysis$.render()
• jmvcore::Analysis$.save()
• jmvcore::Analysis$.savePart()
• jmvcore::Analysis$.setCheckpoint()
• jmvcore::Analysis$.setParent()
• jmvcore::Analysis$.setReadDatasetHeaderSource()
• jmvcore::Analysis$.setReadDatasetSource()
• jmvcore::Analysis$.setResourcesPathSource()
• jmvcore::Analysis$.setStatePathSource()
• jmvcore::Analysis$addAddon()
• jmvcore::Analysis$asProtoBuf()
• jmvcore::Analysis$asSource()
• jmvcore::Analysis$check()
• jmvcore::Analysis$init()
• jmvcore::Analysis$optionsChangedHandler()
• jmvcore::Analysis$postInit()
• jmvcore::Analysis$print()
• jmvcore::Analysis$readDataset()
• jmvcore::Analysis$run()
• jmvcore::Analysis$serialize()
• jmvcore::Analysis$setError()
• jmvcore::Analysis$setStatus()
• jmvcore::Analysis$translate()
• ClinicoPath::patientsimilarityBase$initialize()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

patientsimilarityClass$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
# Discover patient subgroups
patientsimilarity(
  data = clinical_data,
  vars = c("age", "tumor_size", "grade", "ki67"),
  method = "tsne",
  colorBy = "survival_status",
  performClustering = TRUE
)
} # }