Different metabolomics and profiling applications using LC-MS in drug discovery
and development, clinical research, and food science all share the common need to quickly analyze a large number of highly complex samples. Pattern recognition techniques are used to filter the relevant information from different sample groups and thus to detect compounds that differentiate samples.

Reduce Dimensionality while Retaining Information

The ProfileAnalysis™ software is the ideal tool to analyze large numbers of complex LC-MS profiles and extract the relevant information about the distribution of samples and the group characteristics with statistical means like PCA. The goal of PCA is to reduce the dimensionality in a data set while retaining all relevant information. PCA visualizes the distribution of samples in the scores plot and ranks the variables, here RT-m/z-buckets, according to the variance they explain. The preparation of LC-MS analyses for PCA is performed in a simple bucketing step:
RT-m/z-buckets are defined and the corresponding intensities are entered for each sample. As an unsupervised pattern recognition technique, PCA does not require any additional information about samples or tentative groups for calculation. All possible principle components (PCs) are calculated at once, making re-calculations for visualization of higher PCs obsolete. Many more tools for model building like cross validation complete the statistical analysis.
ProfileAnalysis provides the direct interaction of statistical results with the original LC-MS data. This permits the fast identification of statistically significant compounds on-the-fly by unique accurate mass and True Isotopic Pattern Analysis (TIP™) e.g. from micrOTOF™ ESI-TOF, or by MS/MS spectra, e.g. from HCTultra™ quadrupole ion trap or micrOTOF-Q™ ESI-Qq-TOF mass spectrometer.