Package: adverSCarial 1.10.0

Ghislain FIEVET

adverSCarial: adverSCarial, generate and analyze the vulnerability of scRNA-seq classifier to adversarial attacks

adverSCarial is an R Package designed for generating and analyzing the vulnerability of scRNA-seq classifiers to adversarial attacks. The package is versatile and provides a format for integrating any type of classifier. It offers functions for studying and generating two types of attacks, single gene attack and max change attack. The single-gene attack involves making a small modification to the input to alter the classification. The max-change attack involves making a large modification to the input without changing its classification. The CGD attack is based on an estimated gradient descent. against adversarial attacks. The package provides a comprehensive solution for evaluating the robustness of scRNA-seq classifiers against adversarial attacks.

Authors:Ghislain FIEVET [aut, cre], Sébastien HERGALANT [aut]

adverSCarial_1.10.0.tar.gz
adverSCarial_1.10.0.zip(r-4.7)adverSCarial_1.10.0.zip(r-4.6)adverSCarial_1.10.0.zip(r-4.5)
adverSCarial_1.10.0.tgz(r-4.6-any)adverSCarial_1.10.0.tgz(r-4.5-any)
adverSCarial_1.10.0.tar.gz(r-4.7-any)adverSCarial_1.10.0.tar.gz(r-4.6-any)
adverSCarial_1.10.0.tgz(r-4.6-emscripten)
manual.pdf |manual.html
DESCRIPTION |NEWS
card.svg |card.png
adverSCarial/json (API)

# Install 'adverSCarial' in R:
install.packages('adverSCarial', repos = c('https://bioc-release.r-universe.dev', 'https://cloud.r-project.org'))

On BioConductor:adverSCarial-1.11.0(bioc 3.24)adverSCarial-1.10.0(bioc 3.23)

This package does not link to any Github/Gitlab/R-forge repository. No issue tracker or development information is available.

softwaresinglecelltranscriptomicsclassification

5.16 score 24 scripts 16 exports 14 dependencies

Last updated from:fc30cc183d (on RELEASE_3_23). Checks:8 WARNING, 2 OK. Indexed: no.

TargetResultTimeFilesSyslog
bioc-checksWARNING212
linux-devel-x86_64WARNING225
source / vignettesOK248
linux-release-x86_64WARNING247
macos-release-arm64WARNING141
macos-oldrel-arm64WARNING180
windows-develWARNING312
windows-releaseWARNING418
windows-oldrelWARNING360
wasm-releaseOK194

Exports:advCGDadvCharadvGridMinChangeadvListadvMaxChangeadvModificationsadvRandWalkMinChangeadvSingleGeneadvTraininggetSignGenesmatrixFromSCEmaxChangeOverviewMClassifierpredictWithNewValuesceConvertToHGNCsingleGeneOverview

Dependencies:abindBiocGenericsDelayedArraygenericsgtoolsIRangeslatticeMatrixMatrixGenericsmatrixStatsS4ArraysS4VectorsSparseArrayXVector

adverSCarial, generate and analyze the vulnerability of scRNA-seq classifiers to adversarial attacks
Introduction | Jupyter Notebook examples | Installation | Generate an adversarial attack | Single-gene attack | Max-change attack | CGD attack

Last update: 2024-08-13
Started: 2023-04-16

Vign03_adapt_classifiers
Prepare a classifier with CHETAH and scType | CHETAH | Load data | Adapt the classifier | scType

Last update: 2024-07-17
Started: 2023-04-16

Vign02_overView_analysis
Load data | Load loom file | Load cell type annotations | Run vulnerability analysis with singleGeneOverview and maxChangeOverview | Which attack to choose? | Which classifier is more vulnerable to adversarial attacks? | Which modifications to compare

Last update: 2023-08-28
Started: 2023-04-16

Vign04_advRandWalkMinChange
advRandWalkMinChange | Load data | Advanced attacks with advRandWalkMinChange

Last update: 2023-08-28
Started: 2023-04-16

Readme and manuals

Help Manual

Help pageTopics
Implementation of the IDG4C algorithm.advCGD
adverSCarial classadvChar advChar-class
Grid search of min change adversarial attack. Tries each combination on a cluster, given a list of genes and a list of modifications.advGridMinChange
adverSCarial classadvList advList-class
Find a max change adversarial attack. It finds the longer list of genes you can modify on a cluster without changing its classification.advMaxChange
Returns a modified RNA expression DelayedMatrix, or a modified SingleCellExperiment, for a given cluster, for a given modification.advModifications
Random walk search of min change adversarial attack.advRandWalkMinChange
Find a one gene min change adversarial attack list. A one gene min change adversarial attack refers to the modification of a single gene within a cluster, leading to a change in its classification. The function returns a list of genes/new classification.advSingleGene
The function getSignGenes orders the genes by maximizing the significance of the gene to differentiate the clusters and ensures that they represent at most the variations across all possible pairs of clusters.getSignGenes
Returns the RNA expression matrix from a SingleCellExperiment with unique hgnc gene names in columnsmatrixFromSCE
Gives an overview of the susceptibility to max change attacks, for each cell type, for a given list of modifications.maxChangeOverview
Example cell type classifier for pbmc clustered datasets.MClassifier
Returns a classification and an odd value from a RNA expression DelayedMatrix or a SingleCellExperiment object, for given genes, for a given cluster, for a given modification.predictWithNewValue
Returns a SingleCellExperiment object keeping unique HGNC genesceConvertToHGNC
Gives an overview of the susceptibility to single gene attacks, for each cell type, for a given list of modifications.singleGeneOverview