Interactive explorer (Shiny)¶
scMultiBench ships an interactive Shiny dashboard for exploring benchmark results in depth, scIB metric tables, integration/registration quality, feature-selection overlap, classification accuracy, and time/memory profiles, across every method and dataset, without writing any code.
Requirements¶
The dashboard is an R Shiny app built on shinydashboard, plotly, and DT:
What you can explore¶
-
scIB metrics
Batch-correction and bio-conservation sub-scores per method, summarised as the familiar scIB bubble tables and overall rankings.
-
Integration & registration
Side-by-side embeddings and registration quality across vertical, diagonal, mosaic, cross, and spatial integration scenarios.
-
Feature selection
Overlap and correlation of selected features across methods (intersection and rank-correlation views).
-
Classification
Downstream cell-type classification accuracy built on each integrated embedding.
-
Time & memory
Runtime and peak-memory profiles to weigh accuracy against cost.
Source¶
The dashboard is served from the hosted instance above; its source is not part
of the public scMultiBench repository. If you need the numbers behind these
views programmatically, the API reference (mtb.load_results,
mtb.plot.bubble) reproduces them in Python.