Machine learning reveals genes impacting oxidative stress resistance across yeasts

Area of Science:

• Microbiology
• Genetics
• Computational Biology

Background:

• Reactive oxygen species (ROS) are critical molecules impacting yeast survival during metabolism and host interactions.
• Understanding yeast ROS resistance mechanisms is vital for clinical and biotechnological applications.

Purpose of the Study:

• To investigate the variation in ROS resistance across the Saccharomycotina yeast subphylum.
• To identify gene families predictive of ROS resistance using machine learning (ML).
• To guide experimental validation of genetic factors influencing ROS resistance.

Main Methods:

• Characterized ROS resistance variation in Saccharomycotina using tert-butyl hydroperoxide.
• Employed ML to identify predictive gene family sizes for ROS resistance.
• Quantitatively estimated feature contributions for guiding experimental validation.

Main Results:

• Identified cell wall organization and two reductase gene families as highly predictive of ROS resistance.
• Demonstrated that overexpressing the old yellow enzyme (OYE) reductase enhances ROS resistance in Kluyveromyces lactis.
• Showed Saccharomyces cerevisiae mutants lacking mannosyltransferase genes are hypersensitive to ROS.

Conclusions:

• Established a framework for using ML to discover genetic mechanisms of trait variation in yeast.
• Highlighted the role of specific gene families in ROS resistance, informing trait manipulation strategies.
• Provided insights applicable to clinical and biotechnological advancements in yeast research.