Tropism Analysis Package

TAP helps drive discovery of host factors associated with viral tropism in sc/nRNA-seq datasets by automating infectivity clustering, then applying an essemble feature selection pipeline to each subset of cells. The results of each subset are stored and presented in a easy-to-use intuitive interface.

Features

• Explore your data in an interactive user interface.
• Uses an ensemble ML approach to rank important features of tropism.
• Easily share results as an HTML file.
• Simple and configurable parameters.

Installation

We recommend installing in a fresh Python (>=3.12) environment.

pip install scTap

The package is imported in Python as tap:

import tap as t

Live Demo

A curated selection of multiplexed AAV TAPs to explore.
aavdb.com

Usage

TAP accepts an AnnData object or file and generates an interactive HTML report that can be opened in any browser, hosted on a server, or shared. A basic example is shown below. The package expects raw counts in the AnnData object, so the input data should include the original expression values in adata.raw.X for the most reliable results.

import tap as t

parameters = {
    "filename": "PCCM.h5ad",
    "name": "PCCM",
    "genes": ["AAV1", "AAV2", "Retro", "AAV9"],
    "categories": ["cell_type"],
    "outputName": "test.html",
	"clusterThreshold": 4,
    "excludeMarkers": True,
}

results = t.TAP(**parameters)

The published package name is scTap, but the Python import remains tap.

TAP constructor parameters

The TAP class accepts the following constructor arguments. These options control input data handling, clustering, feature selection, plotting, and output generation.

Parameter	Default	Description
name	"Heatmap"	Default: "Heatmap". Use a custom name for the report and related metadata.
filename	None	Default: None. Provide an AnnData file path when you are loading from disk instead of using adataObject.
adataObject	None	Default: None. Pass a preloaded AnnData object here instead of supplying filename.
annSqlDB	None	Default: None. Supply an AnnSQL database identifier to convert into AnnData.
categories	[]	Default: []. Provide one or more metadata columns such as cell_type or serotype for grouping.
categoryNames	[]	Default: []. Use custom display labels for the categories shown in the report.
genes	[]	Default: []. Provide one or more gene names or features to include in the analysis.
exclude	[]	Default: []. Add genes or features to leave out of the analysis.
useRaw	False	Default: False. Set to True to use raw AnnData values when they are available.
outputPath	"."	Default: ".". Set this to a directory where the output HTML and related files should be written.
outputName	"tap.html"	Default: "tap.html". Use a different filename if you want a custom output file name.
cpus	2	Default: 2. Increase this for more parallel work on larger datasets.
mapOnly	False	Default: False. Set to True to generate the map outputs without the full reporting workflow.
showDE	True	Default: True. Set to False to suppress differential expression results.
showRF	True	Default: True. Set to False to skip random forest results.
showUMAP	True	Default: True. Set to False to hide UMAP plots.
showDetails	True	Default: True. Set to False for a more compact report.
clusterMethod	"threshold"	Default: "threshold". Other supported option is "gaussian".
clusterThreshold	1	Default: 1. Increase or decrease this threshold to change the clustering cutoff.
useLog10	False	Default: False. Set to True to log-transform values before downstream analysis.
showPlots	False	Default: False. Set to True to include additional plots in the output.
excludeGenes	[]	Default: []. Add genes to remove before modeling.
deMethod	"wilcoxon"	Default: "wilcoxon". Other common choices include "t-test", "logreg", or other scanpy-supported methods.
rfHyperParameterTune	False	Default: False. Set to True to tune the random forest hyperparameters.
rfHyperParameterIterations	50	Default: 50. Increase this for more tuning coverage.
rfType	"classifier"	Default: "classifier". Other option is "regressor".
balance	None	Default: None. Other options include "smote", "randomoversampler", "adasyn", "borderlinesmote", "kmeanssmote", or "svmsmote".
useAllGenes	False	Default: False. Set to True to use all genes rather than only the input subset.
rfPermuteFeatureImportance	False	Default: False. Set to True to estimate feature importance by permutation.
rfPermuteRepeats	2	Default: 2. Increase this for more stable permutation estimates.
runCellTypist	False	Default: False. Set to True to run CellTypist annotation.
cellTypistModel	"Mouse_Whole_Brain.pkl"	Default: "Mouse_Whole_Brain.pkl". Other CellTypist model names can be supplied here.
cellTypistPlots	False	Default: False. Set to True to generate CellTypist plots.
cellTypistLevels	3	Default: 3. Increase or decrease the number of prediction levels shown.
minify	False	Default: False. Set to True to minify the generated HTML output.
excludeMarkers	False	Default: False. Set to True to exclude marker genes.
removeOutliers	True	Default: True. Set to False to keep outlier cells in the analysis.
minCells	20	Default: 20. Lower or raise this threshold based on dataset size.
minSeroTypeCells	20	Default: 20. Adjust this for stricter or looser serotype group inclusion.
clusterThresholdGreaterThanOrEqual	None	Default: None. Use this to set a lower bound for threshold-based group assignment.
clusterThresholdLessThanOrEqual	None	Default: None. Use this to set an upper bound for threshold-based group assignment.
replicate_mode	False	Default: False. Set to True for replicate-oriented processing.

Commonly used options include genes, categories, outputPath, outputName, excludeMarkers, mapOnly, and runCellTypist.

Citation

Pavan Kenny et al. 2026 Tropism Analysis Package: Interactive Machine Learning Software to Identify Viral Host Factors Through Single-Cell Host-Virus mRNA Profiling. bioRxiv.