Flavour perception is one of the key factors in determining the preference of a consumer for a food product, implying that an overall balance of the aroma compounds is important. In some Swiss cheeses and also other products such as e. g. sourdough bread, malt aroma is desirable. In the present study, however, we worked with a cheese type which is characterised by its buttery, fruity, herbal notes and its ability to melt. Therefore, maltiness was considered as an off-flavour. Using whole genome analysis and model cheese experiments we aimed to understand the influence of the starter culture composition on the formation of the compounds responsible for malt flavour. In order to link these geno- and phenotypes, we took advantage of machine learning algorithms, allowing us to further understand the strain interactions and a possible prediction of the in vitro outcome in silico. Five compounds were previously identified by dynamic headspace-vacuum transfer in trap-gas chromatography-mass spectrometry-olfactometry (DHS-VTT-GC-MS-O) as the main responsible for the malt aroma: 2-methylpropanal, 2-methylbutanal, 3-methylbutanal, 2-methylbutanol and 3-methylbutanol. In a series of experiments using small model cheeses, we determined the concentration of a selection of these compounds at different time points after spiking with their respective precursor compounds. With these measurements, we then trained random forest models using the observed compound concentrations as the response and the underlying genomic information as the features, allowing us to study the importance of each genetic feature and to predict an outcome with new data. With our technique to study malt aroma formation, it may be possible to predict the formation or degradation of 3-methylbutanal in a mixture based on the genome data of the single strains and further confirm the presence/absence of 3-methylbutanal with a simple small scale cheese model. The same approach could also be used for other products such as sourdough bread, spirits and beers, to name a few.