new {GEARS} R Documentation

create a new GEARS model

Description

Usage

new(x, prior, config);

Arguments

x

the gene expression matrix object of the Perturb-seq experiment data, which could be loaded from a csv table file via the geneExpression::load.expr. [as Matrix]

prior

the prior knowledge regulatory network object, which could be created by the bnlearn::as.prior_net. [as PriorNetwork]

config

the hyper parameter configuration of the GEARS model, which could be created via the new("GEARS_opts"). [as GEARSConfig]

Details

GEARS implements of the interface InsilicoPerturbationExperiment, which could be used as the virtual perturbation experiment container for run knockouts/overexpress/knockdown experiments

Authors

biosystem

Value

a new GEARS model object that the gene regulatory graph has been created from the given prior network and expression data, the training sample set should be set via the training_set api at first and then the model can be trained via the train api.

clr value class

Examples

 imports ["GEARS","bnlearn"] from "biosystem";
 imports "geneExpression" from "phenotype_kit";
 
 let exprData = load.expr("./dataset.csv");
 let samples = read.sampleinfo("./experiment.csv");
 let opts = new("GEARS_opts", Activation = "tanh");
 let net = bnlearn::prior_network(
     TF = c(), target_gene = c(), regulation_type = c(), confidence = c(), evidence = c()
 ) |> as.prior_net();
 
 net <- GEARS::new(exprData, net, opts)
 |> training_set(exprData, controls = c(), perturbed = samples)
 |> train()
 ;
 
 make_exports(c( 
    net |> knockouts(c()), 
    net |> overexpress(c()), 
    net |> knockdown(c()) 
 ), dir = "./GNN_result");
[Package GEARS version 1.0.0.0 Index]