RunRosace

R Documentation

Run Rosace on an Assay/AssaySet object

Description

Rosace operates under the assumption that aligned counts are generated by a time-dependent linear function. This function performs inference using stan under a Bayesian framework to derive the functional score (regression coefficient), error term and parameters phi and sigma.

Usage

RunRosace(object, savedir, mc.cores, debug, install, ...)

## S3 method for class 'AssayGrowth'
RunRosace(
  object,
  savedir,
  mc.cores = 4,
  debug = FALSE,
  install = TRUE,
  pos.label,
  ctrl.label,
  ...
)

## S3 method for class 'AssaySetGrowth'
RunRosace(
  object,
  savedir,
  mc.cores = 4,
  debug = FALSE,
  install = TRUE,
  pos.label,
  ctrl.label,
  ...
)

## S3 method for class 'Rosace'
RunRosace(
  object,
  savedir,
  mc.cores = 4,
  debug = FALSE,
  install = TRUE,
  name,
  type,
  pos.col,
  ctrl.col,
  ctrl.name,
  ...
)

Arguments

object	Rosace, Assay/AssaySet (default)
savedir	directory to save the output
mc.cores	integer, number of cores to use for parallel computing
debug	logical, if TRUE, return a list of cmdstanfit and Score object.
install	logical, if TRUE, install or update cmdstanr There's a chance for Github download error if running multiple instances on a server.
...	Additional arguments to be passed to the run method.
pos.label	vector of true position of variants
ctrl.label	vector of whether variant is in the control group (NA if none provided)
name	Name of the object to be analyzed
type	"Assay" or "AssaySet"
pos.col	For Growth screen, the column name for position in the var.data (optional in no_pos mode)
ctrl.col	For Growth screen, optional for control to have one position index
ctrl.name	For Growth screen, optional, the name of the control type

Details

There are three ways to run the model. If no meta-info of the variant is given, the model treats variant independently. If position info is given, the model will have a position hierarchical layer (grouping variants of the same position together). If an additional control (synonymous) label is given, the model will group the control variants together into one position index.

The debug option will return a list of cmdstanfit object and the final Score object. Users familiar with stan could extract more diagnostics and sampling details from the cmdstanfit object. If a high percentage of divergent transitions exist during sampling, users could debug with this option. However, users not familiar with HMC or stan are recommended to report the error either on GitHub or send an email to the team.

Value

A object with Rosace result (Score object) if debug = FALSE