elastic_model(grn,
opts = NULL,
strict = TRUE,
scale = "linear",
propagation = "cascade",
steps = -1,
damping = -1,
knockdown.residual = -1,
log.offset = 0,
clamp = TRUE);
grn
the fitted GeneRegulatoryNetwork object, it is the grn slot of the elastic_net result. [as GeneRegulatoryNetwork]
opts
the perturbation options, a default configuration is used when this argument is not given. [as ElasticNetPerturbationOptions]
strict
in the strict mode an error will be thrown when the target gene of the perturbation is not found in the model; in the non strict mode a warning will be printed and the wildtype expression will be returned with the Undefined flag marked as TRUE. For a batch run of hundreds of the perturbations it is recommended to disable it and to check the MissingGenes field of the result. [as boolean]
scale
the numeric space of the expression matrix that was used to fit the network: linear (default) or log2cpm (i.e. log2(x + 1)). For a log transformed matrix this argument is required, otherwise the knockdown magnitude will be over-estimated by about 2~3 times. [as string]
propagation
cascade (default) propagates the perturbation to the steady state, onehop only keeps the direct effects, delta propagates the change only (recommended for the context conditioned prediction: it is numerically identical to "mutant minus the matched control"). [as string]
steps
the max cascade iteration steps, a value that is less than 1 means "use the default". [as integer]
damping
the damping factor of the cascade iteration, a value that is less than or equals to 0 means "use the default". [as double]
knockdown.residual
the fraction of the wildtype expression that is kept after a knockdown, a negative value means "use the default" (0.2, i.e. about 80% knockdown). [as double]
log.offset
the constant that was subtracted from the log expression values before the network fitting (i.e. the cross gene mean of the sample when the global expression program was removed). It is added back before applying the knockdown/overexpression ratio and subtracted afterwards, so that the perturbation magnitude is computed on the real CPM scale even when the model space contains negative values. Default 0 (no deconfounding). [as double]
env
the R# runtime environment object. [as Environment]