Rotor-Gene ScreenClust HRM Software analyzes HRM data in 4 steps:
- Normalization
- Generation of a residual plot
- Principal component analysis
- Clustering
The software guides the user through all the steps, giving information about any choices that can be made at each step.
HRM performed on the Rotor-Gene cycler produces raw data (*.rex files) that can be further analyzed using Rotor-Gene ScreenClust HRM Software. In the first step in analysis, raw data are normalized by applying curve scaling to a line of best fit so that the highest fluorescence value is equal to 100 and the lowest is equal to zero. Next, the curves are differentiated and a composite median curve is constructed using the median fluorescence of all samples. The melt traces for each sample are subtracted from this composite median curve to draw a residual plot. The individual sample characteristics are extracted by principal component analysis from the residual plot. Principal component analysis is a well-established method of data analysis. However, Rotor-Gene ScreenClust HRM Software is the first software application to apply principal component analysis to HRM data. Principal component analysis highlights similarities and differences in the data and is used to create a cluster plot in supervised or unsupervised mode (see figure " Identification of a class IV SNP"). Clustering (grouping) of data is performed according to allele.
Supervised mode is often used for SNP genotyping, where the genotypes are known. In supervised mode, the user assigns one or more control samples for each cluster and the software classifies (autocalls) all unknown samples to clusters according to their characteristics. The unsupervised mode is used to find new mutations in the data when there is no prior knowledge or only partial knowledge of the genotypes present in the samples. In unsupervised mode, the software calculates the optimum number of clusters by itself. This feature is an excellent tool for the discovery of new polymorphisms.
The result of analysis in both modes is displayed as an easy-to-interpret cluster plot (see figure " I dentification of a class IV SNP"). Statistical probabilities and typicalities are provided to allow easy comparison of results from different experiments. All data and graphs can be conveniently exported in various formats such as JPG, PDF, CSV, or XLS file formats and are summarized in an experimental report.