Automated post-run analysis of arrayed quantitative PCR amplification curves using machine learning.

Area of Science:

• Biotechnology
• Bioinformatics
• Molecular Biology

Background:

• The TaqMan Array Card (TAC) is a high-throughput platform for simultaneous detection of multiple targets in qPCR.
• Manual post-run analysis of TAC data is time-consuming and prone to subjective interpretation.
• Automation of TAC data analysis is needed to improve efficiency and reproducibility.

Purpose of the Study:

• To develop and validate machine learning models for automating the post-run analysis of TAC data.
• To compare the performance of automated analysis against manual analysis by experts.

Main Methods:

• Trained two eXtreme Gradient Boosting (XGBoost) models using 165,214 qPCR amplification curves.
• A classification model predicted amplification presence, and a second model predicted the cycle threshold (Ct) value.
• Validated models using 5-fold cross-validation and external data from 17 laboratory scientists.

Main Results:

• Internal validation showed high accuracy (0.996) for classification and low Mean Absolute Error (MAE) of 0.590 for Ct prediction.
• External validation achieved 0.997 accuracy and 0.611 MAE.
• Automated analysis outperformed manual analysis by 14 out of 17 scientists.

Conclusions:

• Machine learning models can automate highly-arrayed qPCR data analysis with high accuracy.
• This automated approach saves time and enhances reproducibility for TAC users.
• The method is applicable to other high-throughput qPCR platforms.