Data-driven normalization strategies for high-throughput quantitative RT-PCR

Abstract

Background: High-throughput real-time quantitative reverse transcriptase polymerase chain reaction (qPCR) is a widely used technique in experiments where expression patterns of genes are to be profiled. Current stage technology allows the acquisition of profiles for a moderate number of genes (50 to a few thousand), and this number continues to grow. The use of appropriate normalization algorithms for qPCR-based data is therefore a highly important aspect of the data preprocessing pipeline.

Results: We present and evaluate two data-driven normalization methods that directly correct for technical variation and represent robust alternatives to standard housekeeping gene-based approaches. We evaluated the performance of these methods against a single gene housekeeping gene method and our results suggest that quantile normalization performs best. These methods are implemented in freely-available software as an R package qpcrNorm distributed through the Bioconductor project.

Conclusion: The utility of the approaches that we describe can be demonstrated most clearly in situations where standard housekeeping genes are regulated by some experimental condition. For large qPCR-based data sets, our approaches represent robust, data-driven strategies for normalization.

Figure 1

Coefficient of Variation for Different Normalized Data Sets. The CV values for the three different normalization methods on the PMA dataset are represented here in a barchart. The CV for the non-normalized (raw) dataset is included as a reference. The quantile method is associated with the lowest CV, implying the greatest reduction in technical variation in the data.

Figure 2

Exemplar graph to clarify the interpretation of Figure 3. The graph presents a visual pairwise comparison between two normalization algorithms Q₁ and Q₂ on the same data set. For each gene, we calculate the variance of its Q₁-normalized expression profile and its Q₂-normalized expression profile and plot the log₂-ratio of this variance on the y-axis where Y = log₂ [Q₁-normalized: Q₂-normalized]. A gene's log variance ratio is plotted against its expression (mean Ct value) on the x-axis. The regions where the data points fall in the graph give us an indication of which normalization algorithm produces noisier data and whether there is a differential bias in expression for genes most affected by this noise.

Figure 3

Pairwise Comparisons of Different Normalized Data Sets. Pairwise comparisons between the three different normalization methods and the non-normalized dataset. The graphs represent the log variance ratios for each gene versus its average Ct value. The red line is the smoothed lowess curve that captures the overall trend of the data in the plot. The dotted blue line represents horizontal axis. The direction of the ratio is reflected in each individual figure title, e.g. the ratios in Figure 3.3A are constructed by taking the log₂ transformation of the GAPDH-normalized variance divided by the non-normalized variance for each gene. Points below the dotted blue line correspond to those genes where single gene GAPDH normalization has resulted in a greater reduction in variance relative to the variance of these genes in the non-normalized data.