(261b) An Automated Framework for High-Throughput Kinetic Analysis of qRT-PCR Data

We present a novel approach to the analysis of high-throughput gene expression data from quantitative reverse transcription PCR (qRT-PCR), a technique in widespread use for the determination and validation of differential gene expression. The indicator molecule used in qRT-PCR fluoresces only in the presence of double-stranded DNA, thereby enabling the measurement of double-stranded DNA levels at the end of each PCR cycle. Conventional methods of analysis involve an ad-hoc determination of a threshold fluorescence level within the exponential phase of PCR amplification. A cycle threshold (C_T) value is assigned to each reaction based upon when the fluorescence level reaches the threshold, and relative differential expression is determined by comparing C_T values across samples/reactions. This relative comparison neglects the kinetic fluorescence data and assumes equivalent, or nearly equivalent, reaction efficiencies, which can lead to erroneous results.

Several kinetics-based methodologies have been developed recently to utilize the complete fluorescence data from all the reaction cycles in each assay without imposing the aforementioned fundamental assumption of efficiency equivalency. However, these methodologies require manual determination of the exponential phase and arbitrary specification of background signal for each reaction. While these methods are more accurate and more robust than the standard C_T analysis, they are also significantly more time-consuming and cumbersome to perform for each individual reaction in conventional applications. This renders them ineffective, if not infeasible, for high-throughput applications such as Fluidigm's BioMark platform, ABI's TaqMan Custom Array microfluidic cards, or Luminex multiplex systems. To address these issues, we have developed an automated kinetics-based analysis framework for high-throughput qRT-PCR fluorescence data.

Our automated approach is as straightforward as the conventional C_T analysis in its implementation. At the same time, it improves the analysis significantly by providing an inherent quality control assessment for each reaction. The key aspects involve novel heuristics for background detection to automate background subtraction, as well as improved detection of the exponential phase for kinetic model fitting based on the rate of deviation from baseline. We demonstrate the advantages of our approach with several case studies involving experimental qRT-PCR data sets from ongoing projects studying neuronal adaptation to hypertension and alcohol withdrawal. Based on these case studies, we have developed guidelines for the thresholds and other tunable parameters employed in the automated high-throughput analysis. Our framework provides a robust method for determining differential expression from qRT-PCR fluorescence data from any source, enables the use of emerging high-throughput technologies, and improves conventional qRT-PCR data analysis.