Predictive modeling and network analysis uncover novel regulators of multi-stress tolerance in Saccharomyces...

Understanding molecular responses to multi-stress conditions in Saccharomyces cerevisiae is crucial for optimizing industrial strains and enhancing stress tolerance. This study uses machine learning and data mining approaches to unravel the critical genes by which Saccharomyces cerevisiae responds to integrated stresses.

RNA-seq analysis was performed on 48 sequencing runs; the subsequent results were analyzed by Weighted Gene Co-expression Network Analysis, and a distinct red module consisting of 371 genes associated with stress response was identified. The Boruta feature selection algorithm was applied to this module, confirming 56 important genes while classifying others as tentative or unimportant. Subsequent Random Forest analysis refined this set by highlighting essential genes, achieving robust classification accuracy, and focusing intensely on the most relevant genes for stress adaptation. Our integrative computational approach, including Module Membership and Gene Significance tests and machine learning methods, identified known and unknown candidate genes linked to oxidative stress, osmotic stress, and glucose deprivation. Notably, PPN1, SAM3, PLN1, Icl1, CTA1, AMS1, GAC1, and ADH2 genes demonstrated significant roles in adaptation mechanisms under osmotic, oxidative, and acid stress.

This preliminary study clarified the underlying molecular network responses to complex stress conditions in yeast, which can be applied for further efficient strain engineering to improve yeast robustness for industrial applications.