No control genes required: Bayesian analysis of qRT-PCR data.

Area of Science:

• Molecular Biology
• Bioinformatics
• Statistical Genetics

Background:

• Traditional quantitative reverse-transcription PCR (qRT-PCR) analysis methods struggle with high sampling variances of low-abundant targets.
• Existing model-based approaches lack a natural way to integrate assumptions about control gene stability into the model-fitting process.

Purpose of the Study:

• To develop a more powerful and versatile model-based analysis for qRT-PCR data.
• To address limitations in handling low-abundance targets and control gene stability in existing methods.

Main Methods:

• Representing raw qRT-PCR data as molecule counts and employing generalized linear mixed models under Poisson-lognormal error.
• Utilizing a Markov Chain Monte Carlo (MCMC) algorithm to estimate experimental factor effects on gene expression.
• Incorporating prior knowledge of control gene stability directly into the model or performing analysis without such assumptions.

Main Results:

• The Poisson-based model correctly specifies the mean-variance relationship of PCR amplification and utilizes zero-count data.
• The method offers flexibility in control gene usage, providing sensible results even without prior stability assumptions or control genes.
• A Bayesian analogue to classic analysis is presented, estimating all gene expression changes jointly within a single model.

Conclusions:

• The developed methodology extends relative quantification to low-abundance targets and allows analysis without target stability assumptions.
• This novel approach offers enhanced flexibility and power compared to standard delta-delta Ct analysis.
• The procedures are implemented in the MCMC.qpcr package for R.