fit_dropout_model: Fit a Bayesian model of dropout CRISPR screen data
In andrewGhazi/malacoda: Bayesian Analysis of High-Throughput Genomic Assays

Description Usage Arguments Details Value Note Examples

This function fits a Bayesian model of survival/dropout CRISPR screen data. It uses a negative binomial to model the input and output counts of gRNAs, adjusting the results appropriately to account for sequencing depth.

fit_dropout_model(
  dropout_data,
  out_dir = NULL,
  n_cores = 1,
  tot_samp = 10000,
  n_warmup = 500,
  n_chains = 4,
  rep_cutoff = 0.1,
  plot_rep_cutoff = TRUE,
  verbose = TRUE
)

`dropout_data`	a data frame of dropout data. See details for column requirements.
`out_dir`	path to output directory
`n_cores`	number of cores to utilize
`tot_samp`	total number of MCMC draws to take, spread evenly across chains
`n_warmup`	total number of warmup draws to take from each MCMC chain
`n_chains`	number of MCMC chains to run in each sampler
`rep_cutoff`	a representation cutoff quantile (0 to 1)
`plot_rep_cutoff`	logical indicating whether to plot the representation cutoff histogram
`verbose`	logical indicating whether to print messages

dropout_data requires the following columns:

gene_id - character column giving a unique identifier for each gene
gRNA - character column giving identifiers for individual gRNAs (usually the gRNA sequence itself)
input count columns - columns of sequencing counts of the input gRNA library. Multiple columns for sequencing replicates are allowed (which require unique identifiers). Column names must contain the string "input".
output count columns - columns of sequencing counts of gRNAs in the output libraries. Multiple columns allowed (which in turn require unique names). Column name must contain the string "output".

a data frame of input counts, fit and model statistics for the log-fold-change for each input gene.

Currently this function only supports marginal priors. If you want to use grouped/conditional priors, contact the malacoda developers.

The gene_data column in the output contains only the gRNAs that passed the representation cutoff.

# This example uses too-few MCMC samples for the sake of run time. Convergence will be poor.

fit_dropout_model(dropout_data = dropout_example,
n_cores = 1,
tot_samp = 20,
n_warmup = 5)