Assess concordance between tables of observed and predicted phenotypes

Calculates concordance from two data frames in standard phenotype table format, one with observed phenotype data and the other with predicted phenotype calls (e.g. output from AMRrules).

Usage

concordance_from_tables(
  pheno_table,
  pheno_pred_table,
  true_SIR_col = "pheno_eucast",
  true_ecoff_col = "ecoff",
  pred_SIR = "clinical category",
  pred_ecoff = "phenotype",
  sample_col = "id",
  drug_col = "drug",
  measure_col = "mic"
)

Arguments

pheno_table

A tibble or data frame containing real (observed) phenotype data, in the format output by import_pheno().

pheno_pred_table

A tibble or data frame containing predicted phenotype calls, e.g. AMRrules predictions imported in AMRgen using import_amrrules_predictions().

true_SIR_col

Character. Name of the column containing S/I/R calls interpreted from real data against clinical breakpoints (default "pheno_eucast").

true_ecoff_col

Character. Name of the column containing WT/NWT calls interpreted from real data against ECOFF (default "ecoff").

pred_SIR

Character. Name of the column containing S/I/R calls predicted from genotypes (default "clinical category").

pred_ecoff

Character. Name of the column containing WT/NWT calls predicted from genotypes (default "phenotype").

sample_col

Character. Name of the column containing sample identifiers. This must be the same in both tables (default "id").

drug_col

Character. Name of the column containing drug agent identifiers. This must be the same in both tables (default "drug").

measure_col

Character. Name of the column containing observed MIC or disk measurements for plotting. Valid options are "mic" or "disk" (default "mic").

Value

A named list with the following elements:

• obs_pred: A copy of the pheno_table with the predictions from pheno_pred_table merged in.

• metrics: A tibble listing concordance metrics, comparing observed vs predicted R and/or NWT calls, for all drugs with matched samples in both input tables.

• drugs_R: A long-form tibble listing the number of samples with each combination of observed and predicted binary calls for R, for all drugs with matched samples in both input tables.

• drugs_NWT: A long-form tibble listing the number of samples with each combination of observed and predicted binary calls for NWT, for all drugs with matched samples in both input tables.

• plot_R: A list containing two-panel plots summarising observed vs predicted clinical categories, one for each drug.

• plot_NWT: A list containing two-panel plots summarising observed vs predicted WT/NWT categories, one for each drug.

• plot_R_dist: A list containing plots of the assay value distributions (MIC or disk zones), coloured by S/I/R prediction, one for each drug.

• plot_NWT_dist: A list containing plots of the assay value distributions (MIC or disk zones), coloured by WT/NWT prediction, one for each drug.

Examples

if (FALSE) { # \dontrun{
concordance_from_tables(true_phenotypes, predicted_phenotypes)
} # }