Fundamentals for the Automatic Classification of Quantitative PCR AmplificationCurves – A Biostatistical Approach

Stefan Rödiger1, Andrej-Nikolai Spiess2, Michał Burdukiewicz3
1) Brandenburg University of Technology Cottbus – Senftenberg, Germany;
2) University Medical Center Hamburg-Eppendorf, Germany;
3) Warsaw University of Technology, Poland;

Quantitative polymerase chain reaction (qPCR) is a widely used bioanalytical method in forensics, human diagnostics and life sciences. With this method nucleic acids are detected and quantified. In qPCRs, the enzymatic amplification of the target DNA (amplicon) is monitored in real-time by fluorescent reporter molecules marking the synthesized PCR products cycle by cycle. The measured fluorescence is proportional to the amplicon amount.
For diagnostic and forensic applications in particular, the question arises for example as to whether an amplification reaction is negative or positive. Of interest is also and automatic classification of the quality of amplification curves. Until now, such classification was usually performed manually or on the basis of fixed threshold values. However, this approach is error-prone if inadequate thresholds are used or the user performs the classification subjectively based on his experience.
Therefore, the classifications of the same sample may not be identical for different users. Such errors are problematic because they can lead to an erroneous judgement. Therefore we developmed a scientific open source software, called PCRedux ( With this software, predictors (features) of amplification curves can be calculated automatically. A predictor is a quantifiable informative property of an amplification curve. A set of statistical algorithms for the calculation of predictors os proposed. The work also shows how predictors can be used in tests and logical combinations to perform machine-based classifications.
All scientific work depends on the data, with open data in particular being regarded as a cornerstone of science. Since no data sets of classified amplification curves were available, the work also deals with the aggregation, management and distribution of classified qPCR data sets. Manual classification of amplification curves is time-consuming and error-prone, especially for large data sets. To improve this, auxiliary tools have been developed.
A open approach for curve-shape based group classification was proposed.

Back to GQ2019 overview page
Bookmark the permalink.

Comments are closed.