The potential of MALDI-TOF profiling for predicting potential applications of yeast strains in the beverage sector was assessed. A panel of 59 commercial yeasts (47 wine and 12 brewing yeasts) was used to validate the concept whereby 2 culture media (YPD agar and YPD broth), as well as two mass ranges m/z 500-4000 and m/z 2000-20,000, were evaluated for the best fit. Three machine learning-based algorithms, PCA, MDS, and UMAP, in addition to a hierarchical clustering method, were employed. Profiles derived from broth cultures yielded more peaks, but these were less well-defined compared with those from agar cultures. Hierarchical clustering more clearly resolved different species and gave a broad overview of potential strain utility, but more nuanced insights were provided by MDS and UMAP analyses. PCA-based displays were less informative. The potential of MALDI-TOF proteomics in predicting the utility of yeast strains of commercial benefit is supported in this study, provided appropriate approaches are used for data generation and analysis.
Keywords: MALDI-TOF analysis; UMAP; brewing yeast; commercial wine yeast; winemaking.