ABC inference of BFB simulation parameters

Usage

abc_inference(
  cna_data,
  allele,
  chromosome,
  n_simulations = 10000,
  tolerance_quantile = 0.01,
  n_cores = 1,
  pos = NULL,
  bin_length = NULL
)

Arguments

cna_data

Data frame with columns cell_id, chr, start, end, CN, A, B.

allele

Allele to match ("A", "B", or "CN").

chromosome

Chromosome to focus on (character, e.g. "7").

n_simulations

Total number of simulations to run. Default: 10000.

tolerance_quantile

Fraction of simulations to accept. Default: 0.01.

n_cores

Number of parallel cores (uses parallel::mclapply when > 1). Default: 1.

pos

Optional integer bin index for the BFB hotspot in the simulation. Automatically derived from the observed data as the bin with the highest CN variance if NULL (default).

bin_length

Optional bin size in bp. Inferred from observed data if NULL.

Value

A list containing:

accepted_params

Data frame of accepted parameter draws.

param_summary

Per-parameter mean / median / SD / 95\ n_simulationsTotal simulations attempted. n_acceptedNumber of accepted simulations. acceptance_rateFraction accepted out of valid (non-error) simulations. tolerance_thresholdDistance cutoff used for acceptance.

Uses rejection ABC to infer simulation parameters that reproduce the copy number distribution of an observed single-cell dataset. The simulator (bridge_sim) is run n_simulations times with parameters drawn from sample_priors(). The tolerance_quantile fraction with the smallest distance to the observed summary statistics is retained as the approximate posterior.