Estimating the accuracy of polymerase chain reaction-based tests using endpoint dilution

Area of Science:

• Molecular Biology
• Biostatistics
• Medical Diagnostics

Background:

• Polymerase Chain Reaction (PCR) tests are widely used for detecting microorganisms and DNA sequences.
• The definition and measurement of "sensitivity" in PCR assays are often imprecise in existing literature.
• Accurate quantification of sensitivity is essential for reliable diagnostic outcomes.

Purpose of the Study:

• To propose a rigorous definition of PCR test sensitivity as a function of target DNA molecules.
• To develop and compare statistical models for estimating the "sensitivity curve" of PCR assays.
• To evaluate these models using a novel test for Mycoplasma genitalium.

Main Methods:

• Utilizing serially diluted replicate samples with known target DNA concentrations.
• Developing parametric, nonparametric, and semiparametric (spline-based) models to estimate the sensitivity curve.
• Disentangling random variations in target molecule counts from intrinsic test sensitivity.

Main Results:

• Demonstrated a feasible method to estimate the "sensitivity curve" by analyzing replicate aliquots of serially diluted DNA.
• Successfully applied and compared different modeling approaches (parametric, nonparametric, spline-based).
• Provided a more precise estimation of sensitivity for a new Mycoplasma genitalium test.

Conclusions:

• The proposed functional definition of sensitivity improves the understanding and quantification of PCR test performance.
• The developed statistical models offer robust methods for estimating PCR sensitivity, applicable to various assays.
• Accurate sensitivity estimation is vital for the clinical interpretation of PCR-based diagnostic tests.