Algorithms for automated detection of hook effect-bearing amplification curves

TY - JOUR

T1 - Algorithms for automated detection of hook effect-bearing amplification curves

AU - Burdukiewicz, Michał

AU - Spiess, Andrej-Nikolai

AU - Blagodatskikh, Konstantin A

AU - Lehmann, Werner

AU - Schierack, Peter

AU - Rödiger, Stefan

PY - 2018/12

Y1 - 2018/12

N2 - Amplification curves from quantitative Real-Time PCR experiments typically exhibit a sigmoidal shape. They can roughly be divided into a ground or baseline phase, an exponential amplification phase, a linear phase and finally a plateau phase, where in the latter, the PCR product concentration no longer increases. Nevertheless, in some cases the plateau phase displays a negative trend, e.g. in hydrolysis probe assays. This cycle-to-cycle fluorescence decrease is commonly referred to in the literature as the hook effect. Other detection chemistries also exhibit this negative trend, however the underlying molecular mechanisms are different. In this study we present two approaches to automatically detect hook effect-like curvatures based on linear (hookreg) and nonlinear regression (hookregNL). As the hook effect is typical for qPCR data, both algorithms can be employed for the automated identification of regular structured qPCR curves. Therefore, our algorithms streamline quality control, but can also be used for assay optimization or machine learning.

AB - Amplification curves from quantitative Real-Time PCR experiments typically exhibit a sigmoidal shape. They can roughly be divided into a ground or baseline phase, an exponential amplification phase, a linear phase and finally a plateau phase, where in the latter, the PCR product concentration no longer increases. Nevertheless, in some cases the plateau phase displays a negative trend, e.g. in hydrolysis probe assays. This cycle-to-cycle fluorescence decrease is commonly referred to in the literature as the hook effect. Other detection chemistries also exhibit this negative trend, however the underlying molecular mechanisms are different. In this study we present two approaches to automatically detect hook effect-like curvatures based on linear (hookreg) and nonlinear regression (hookregNL). As the hook effect is typical for qPCR data, both algorithms can be employed for the automated identification of regular structured qPCR curves. Therefore, our algorithms streamline quality control, but can also be used for assay optimization or machine learning.

KW - Journal Article

U2 - 10.1016/j.bdq.2018.08.001

DO - 10.1016/j.bdq.2018.08.001

M3 - SCORING: Journal article

C2 - 30560061

VL - 16

SP - 1

EP - 4

JO - Biomol Detect Quantif

JF - Biomol Detect Quantif

SN - 2214-7535

IS - 16

ER -