Abstract

Saccharomyces cerevisiae is a model organism with key industrial relevance in wine fermentation and other biotechnological processes. Identifying single-nucleotide polymorphisms (SNPs) linked to quantitative traits is essential for understanding the genetic basis of strain performance. Most previous studies have frequently focused on lineage-based comparisons, limiting the resolution of genotype-phenotype associations. To deal with this constraint, we applied regression and machine-learning models to characterise SNPs and genes associated with continuous oenological traits in wine strains. Four phenotypes were evaluated: proliferation efficiency, proliferation rate, lag phase, and area under the growth curve. Genome-wide SNP data were encoded and used to train predictive models based on Ridge regression, Support Vector Machines, K-nearest Neighbours, and Random Forest. Model performance was evaluated through cross-validation, and key SNPs were prioritised using feature-importance metrics. Functional enrichment of the associated genes was performed using curated biological databases to provide a mechanistic context. This integrative approach identified genomic regions consistently linked to fermentation traits and revealed candidate genes involved in stress adaptation and metabolic regulation. The results refine genotype-phenotype associations in wine yeasts and establish a reproducible framework for future experimental validation of functionally relevant variants.