Calibration using constrained smoothing with applications to mass spectrometry data.

Area of Science:

• Biotechnology
• Analytical Chemistry
• Proteomics

Background:

• Linear regressions are standard for mass spectrometry calibration but often fail due to non-linear and non-monotone proteomic data.
• Mass spectrometry data frequently exhibit heteroscedasticity, with increasing variance near dynamic range boundaries.
• Existing spline methods improve calibration but do not fully address heteroscedasticity and range limitations.

Purpose of the Study:

• To develop a robust functional data approach for mass spectrometry calibration curves and limits of quantitation.
• To address limitations of linear regression and spline methods in proteomic analysis.
• To define the limit of detection within the new calibration paradigm.

Main Methods:

• Utilized a functional data approach assuming bounded, convex variance and a monotone calibration curve within the limits of quantitation.
• Employed an iterative smoothing method to simultaneously account for variance and monotonicity restrictions.
• Defined the limit of detection based on signal presence without accurate measurement.

Main Results:

• The proposed method accurately defines calibration curves and limits of quantitation, outperforming linear regression and spline techniques.
• Achieved optimal convergence rates under weak conditions, demonstrating computational efficiency.
• Successfully applied to real-world proteomic data, validating its effectiveness.

Conclusions:

• The functional data approach provides a more accurate and reliable calibration strategy for mass spectrometry.
• This methodology enhances the precision of proteomic quantification and the determination of analytical limits.
• The approach is adaptable to other bounded restrictions beyond convexity.