PCRedux: A Quantitative PCR Machine Learning Toolkit
- 1. Autonomous University of Barcelona, Bellaterra, Spain; Medical University of Białystok, Białystok, Poland
- 2. Soilytix GmbH, Hamburg, Germany
- 3. Warsaw University of Technology, Warsaw, Poland
- 4. Pirogov Russian National Research Medical University, Moscow, Russia
- 5. BTU Cottbus–Senftenberg, Faculty of Health Brandenburg, Senftenberg, Germany; Cottbus–Senftenberg, Faculty Environment and Natural Sciences, Senftenberg, Germany
qPCR (quantitative polymerase chain reaction) is indispensable in research, diagnostics and forensics, because it provides quantitative information about the amount of DNA in a sample . The interpretation of amplification curves (ACs) is often difficult if the curve does not follow a typical sigmoidal trajectory.
PCRedux is an R package for feature extraction and classification in the realm of explainable machine learning, which uses statistical functions to compute 90 boolean and numerical descriptors from ACs. It can also be used to determine Cq values and amplification efficiencies (E) for high-throughput analysis.
Given the lack of class-labeled qPCR data sets, PCRedux includes functions for aggregation, management and dissemination of qPCR datasets that can, but must not necessarily be, trichotomously classified into negative, positive and ambiguous curves.