# A proposal to improve calibration and outlier detection in high-throughput mass spectrometry

## TL;DR

The paper discusses how to improve calibration and detect outliers in mass spectrometry to ensure accurate results.

## Contribution

The paper provides a post-hoc explanation for calibration requirements and recommends better methods for curve building and outlier detection.

## Key findings

- Using seven calibration points ensures a 95% confidence limit in regression.
- Quadratic or cubic fits increase the risk of poor model fitting.
- Serial dilution causes leverage issues at the upper range of measurements.

## Abstract

•Seven calibration points implies a 95% confidence limit on the regression.•Most labs use calibration to map the detector instead of confirm instrument stability.•Quadratic or cubic fits to the data increase the risk of a poorly fit model.•Serial dilution creates a leverage problem for the upper end of the range.

Seven calibration points implies a 95% confidence limit on the regression.

Most labs use calibration to map the detector instead of confirm instrument stability.

Quadratic or cubic fits to the data increase the risk of a poorly fit model.

Serial dilution creates a leverage problem for the upper end of the range.

Instrument calibration, required for any accurate quantitative calculation, is a trivial process when performed correctly, but is also full of easily overlooked stumbling blocks. To minimize the risk of error associated with improper calibrations, national and international guidance dictates a minimum number of calibrators and the threshold at which a measurement becomes an outlier. Evidence from industry practice, which conflicts with regulatory guidance, suggests that most groups are focused on remapping their detector with each run. We present a post facto explanation for the calibrator minimum and provide recommendations for curve building, which include improved outlier detection for high-volume mass spectrometry laboratories.