Performance evaluation of tile-based Fisher Ratio analysis using a benchmark yeast metabolome dataset

• Discovery based analysis was applied to a benchmark yeast metabolome dataset.
• Tile-based Fisher Ratio analysis was validated against a previously analyzed dataset.
• Null distribution analysis was utilized for threshold determination to optimize data mining.
• 94 up-/down-regulated metabolites were identified above the threshold.
• The analysis was replicated four times and all replicates were in excellent agreement with the benchmark dataset.

Performance of tile-based Fisher Ratio (F-ratio) data analysis, recently developed for discovery-based studies using comprehensive two-dimensional gas chromatography coupled with time-of-flight mass spectrometry (GC × GC–TOFMS), is evaluated with a metabolomics dataset that had been previously analyzed in great detail, but while taking a brute force approach. The previously analyzed data (referred to herein as the benchmark dataset) were intracellular extracts from Saccharomyces cerevisiae (yeast), either metabolizing glucose (repressed) or ethanol (derepressed), which define the two classes in the discovery-based analysis to find metabolites that are statistically different in concentration between the two classes. Beneficially, this previously analyzed dataset provides a concrete means to validate the tile-based F-ratio software. Herein, we demonstrate and validate the significant benefits of applying tile-based F-ratio analysis. The yeast metabolomics data are analyzed more rapidly in about one week versus one year for the prior studies with this dataset. Furthermore, a null distribution analysis is implemented to statistically determine an adequate F-ratio threshold, whereby the variables with F-ratio values below the threshold can be ignored as not class distinguishing, which provides the analyst with confidence when analyzing the hit table. Forty-six of the fifty-four benchmarked changing metabolites were discovered by the new methodology while consistently excluding all but one of the benchmarked nineteen false positive metabolites previously identified.