Evaluation and applicability of advanced/exotic qPCR quantification strategies and their implementation

Abstract
Over its now three years of existence and development, the ‘qpcR’ package for the R statistical environment has matured into a quantitative real-time PCR (qPCR) data analysis software that houses nearly 70 functions and example datasets assisting in qPCR analysis. The versatile R statistical language makes it relatively effortless to implement published algorithms or new ideas in a short time, and test these new algortihms by the 20 included datasets derived from many different PCR platforms. The package includes in its current version (1.3-4) many ‘replications’ of widely-used methods (i.e. window-of-linearity, REST approach, exponential fitting, sigmoidal fitting, threshold cycle estimation by second derivative maximum, Cy0, comparative quantitation, ratio calculation in a batch format, etc). All of these methods have been compared with the original versions, as to ensure result compatibility/identity. Many implemented functions have also been officially approved by the original authors. The talk will focus on somewhat exotic implementations of qPCR quantitation that are relatively new. While some of them are already published (but not necessarily available in a software), we will introduce methods that are still under development or amenable to public discussion in respect to their general applicability within the qPCR community. The methods that will be presented with short examples are:
1) Monte-Carlo simulation of error propagation (short).
2) Bootstrapping qPCR fits.
3) The development of a tilted threshold, which gives significantly better calibration values from high dilution line by an iterative process.
4) A six-parameter sigmoidal model which models the baseline slope accurately.
5) The MAK2/MAK3 mechanistic models from Boggy et al.
6) The maxRatio method from Shain et al.
7) Several methods for automatic tagging of unsuccessful runs.
8) Using the PRESS statistic for goodness-of-fit.
9) The exceptional speed of R when analyzing high-throughput data. We show this this on a dataset with 9000 runs (raw fluorescence data), whose import, model fitting, quality control (failed runs detection), threshold cycle and efficiency calculation takes about 10 minutes.
10) The importance of weighting when fitting nonlinear models to qPCR data.
Some these methods will be presented as short animations, to clarify things. We will also show the feasibility of the methods by testing them on a cohort of datasets.