Calculate external calibration curve results

Calibration curves are calculated for each feature using ISTD-normalized intensities and the corresponding concentrations of calibration samples, as defined in the qc_concentrations metadata. The regression fit model (linear or quadratic) and the weighting method (either "none", "1/x", or "1/x^2") can be defined globally via the arguments fit_model and fit_weighting for all features, if fit_overwrite is TRUE. Alternatively, the model and weighting can be defined individually for each feature in the feature metadata (columns curve_fit_model and fit_weighting). If these details are missing in the metadata, the default values provided via fit_model and fit_weighting will be used.

Usage

calc_calibration_results(
  data = NULL,
  variable = "feature_norm_intensity",
  include_qualifier = TRUE,
  fit_overwrite,
  fit_model,
  fit_weighting,
  ignore_missing_annotation = FALSE,
  include_fit_object = FALSE,
  lod_sigma = c("residual", "intercept")
)

Arguments

data

A MRMhubExperiment object containing the data to be used for calibration.

variable

A character string specifying the variable for calibration. Use "feature_norm_intensity" for typical scenarios involving internal standardization. When performing only external standardization, without internal standardization, use "feature_intensity".

include_qualifier

A logical value. If TRUE, the function will include qualifier features in the calibration curve calculations.

fit_overwrite

If TRUE, the function will use the provided fit_model and fit_weighting values for all analytes and ignore any fit method and weighting settings defined in the metadata.

fit_model

A character string specifying the default regression fit method to use for the calibration curve. Must be one of "linear" or "quadratic". This method will be applied if no specific fit method is defined for a feature in the metadata, or when fit_overwrite = TRUE.

fit_weighting

A character string specifying the default weighting method for the regression points in the calibration curve. Must be one of "none", "1/x", or "1/x^2". This method will be applied if no specific weighting method is defined for a feature in the metadata, or when fit_overwrite = TRUE.

ignore_missing_annotation

If FALSE, an error will be raised if any of the following information is missing: calibration curve data, ISTD mix volume, and sample amounts for any feature.

include_fit_object

If TRUE, the function will return the full regression fit objects for each feature in the metrics_calibration table.

lod_sigma

A character string selecting the response standard deviation (sigma) used in the ICH Q2 LoD/LoQ formulas. Must be one of "residual" (the residual standard error of the regression, Sy/x; the default) or "intercept" (the standard error of the intercept). No averaging of the two is performed.

Value

A modified MRMhubExperiment object with an updated metrics_calibration table containing the calibration curve results, including concentrations, LoD, and LoQ values for each feature.

Details

Additionally, the limit of detection (LoD) and limit of quantification (LoQ) are calculated for each feature based on the calibration curve, following the ICH Q2(R1/R2) approach (LoD = 3.3 sigma / S, LoQ = 10 sigma / S). Here S is the slope of the calibration curve, taken at zero concentration (the linear coefficient coef_b). For a quadratic fit the true slope is b + 2 c x, which reduces to coef_b at zero concentration, so the quadratic term does not contribute to the slope used here. ICH Q2 specifies the slope formula for linear responses only; using the low-concentration tangent slope for a quadratic fit is an approximation beyond the guideline (adequate when the curvature near zero is small).

The response standard deviation sigma is selected via lod_sigma, following the ICH-acceptable choices: "residual" uses the residual standard error of the regression (Sy/x; the default and prior behaviour), while "intercept" uses the standard error of the intercept. The sigma column reported in metrics_calibration is always the residual standard error, independent of this choice.

The results of the regression and the calculated LoD and LoQ values are stored in the metrics_calibration table of the returned MRMhubExperiment object.

References

ICH Harmonised Tripartite Guideline. Validation of Analytical Procedures: Text and Methodology Q2(R1) (2005); Q2(R2) (2023). International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use.

See also

quantify_by_calibration() for calculating concentrations based on external calibration curves.