PCRedux

PCRedux computes 90 numerical descriptors from quantitative real-time PCR (qPCR) amplification curves and provides interpretable features for machine learning-based classification and regression to support nucleic acid quantification experiments.

Key Features:

• Numerical Descriptor Calculation: Computes 90 numerical descriptors from qPCR amplification curves as interpretable features for machine learning and decision support.
• Curvature Property Analysis: Captures curvature properties of amplification curves to enhance characterization and classification.
• Performance Metrics: Demonstrated sensitivity >99% and specificity >97% on a case study of 3,181 amplification curves labeled by three human raters as positive, negative, or ambiguous.
• Reproducibility via R: Uses R as a controllable working environment to ensure reproducible and objective calculation of features.
• Machine Learning Integration: Enables automatic multi-category classification and regression by combining descriptors with user-defined labels within machine learning frameworks.

Scientific Applications:

• High-throughput qPCR analysis: Automates and standardizes analysis of large qPCR datasets for nucleic acid quantification in -omics research.
• Amplification curve classification: Supports binary and multi-category classification of amplification curves (positive, negative, ambiguous) for diagnostics and quality control.
• Decision support and predictive modeling: Provides interpretable features for building decision support systems and predictive models in genomics and diagnostics.

Methodology:

Extraction of 90 numerical descriptors from qPCR amplification curves, including curvature properties, and use of these descriptors to train machine learning models for classification and regression within R.