Researchers use technology to better understand and predict drug-resistant fungi
By Tatum Lyles Flick, Evolutionary Studies communications consultant
Many fungi are naturally resistant to the drugs available to treat them, while others evolve resistance to drugs prescribed in clinical settings. Determining which medication will be most effective in treating a patient can be a time-consuming challenge. 五一茶馆儿 researchers recently found a way to leverage machine learning to find the weak link in a pathogen鈥檚 genetic armor, enabling them to forecast what species might be resistant to which drugs, with an accuracy range between 54 and 75 percent.
The new paper 鈥,鈥 published in PLOS Genetics was co-first authored by former Rokas Lab Ph.D. student Marie-Claire Harrison and current research assistant professor David Rinker. The research group also includes other alumni and colleagues at the University of Wisconsin. The team worked with Saccharomycotina yeasts and eight key antifungal drugs to understand how genomic, ecological, and phenotypic traits shape resistance. They used machine learning to quickly assess the genetic makeup of 532 Saccharomycotina yeasts to better understand what genetic variables might predict resistance in wild populations.
鈥淭丑别 Saccharomycotina subphylum that we analyzed is famously described (well, at least it’s a famous description among yeast biologists) to span an evolutionary distance equivalent to that between humans and sponges,鈥 Rinker said. 鈥淭his can make comparative biology challenging, but just a handful of genomic features resulted in surprisingly good predictive accuracies.鈥
According to the research, genes affecting yeast cell wall composition and colony makeup informed the assessment more than those already known to promote resistance, suggesting 鈥渢hat machine learning can pick up on features that impact resistance, even if their molecular mechanism(s) is indirect.鈥
Rinker was surprised by the effectiveness and accuracy of machine learning models to predict drug resistance across a species with such genomic, geographic, and habitat diversity.
Though scientists have a good understanding of how resistance works in a clinical setting, the team wanted to understand how resistance evolves in natural populations and where it might show up in the genetic code.
鈥淢ost studies that focus on drug resistance never look beyond the clinic or to other related species,鈥 he said. 鈥淭hat means that we are lacking broad, contextual information about how such resistances might evolve and persist. Analyzing this type of data is crucial to better understand how resistance evolves in the clinic.鈥
According to the paper, some wild yeast species have evolved resistance to antifungal medications used since the late 1970s. These drugs can enter the environment through wastewater, exposing more than just medically targeted yeast. Resistance in these wild relatives can occur through different genetic mechanisms than in clinical settings, sometimes affected by fungicide concentration, exposure duration, and the number of fungal species affected.
This wider view of how numerous Saccharomycotina species聽endure in the face of antifungal medications opens new understanding into the novel genetic patterns that allow yeast to carry on and how they might evolve to defend themselves.
鈥淲hile there are different ways to achieve the same functional ends, certain conditions require more extreme fitness tradeoffs on the part of the pathogen in order to adapt rapidly,鈥 Rinker explained. 鈥淏iophysical modeling of fitness effects showed that clinically-evolved variants are predicted to be much more detrimental to the drug’s target than are variants observed in natural isolates.鈥
Healthcare practitioners and pharmaceutical companies could benefit from the ability to predict the probability of drug resistance based on genetic data; however, because drug resistance can result from more than one set of genes, future studies could reveal more useful information.
鈥淥ne next step would be to account for this polygenicity using more sophisticated machine learning encoding strategies,鈥 Rinker said. 鈥淎nother direction to explore would be to see how well these models perform not just across species (as we have presently done) but across hundreds or thousands of samples of the same species.鈥
Many variables affect genetic changes and how drug resistance appears. This complex balance between different yeast genetic makeups and environmental pressures, paired with machine learning, can open doors to new understandings of how to fight infection and reduce resistance. Looking beyond the clinical setting adds substantial data to what researchers know and may offer more insight as additional yeast species are studied.
Citation: Harrison M-C, Rinker DC, LaBella AL, Opulente DA, Wolters JF, Zhou X, Shen XX, Groenwald M, Hittinger CT, Rokas A. (2026) Machine learning identifies novel signatures of antifungal drug resistance in Saccharomycotina yeasts. PLoS Genet 22(3): e1012091. https://doi.org/10.1371/journal.pgen.1012091
Funding Statement: This project was supported by the National Science Foundation under Grants No. DEB-2110403 (C.T.H.); DEB-2110404 (A.R.); in part by the Great Lakes Bioenergy Research Center, U.S. Department of Energy, Office of Science, Biological and Environmental Research Program under Award Number DESC0018409 (C.T.H.); and the National Institute of Food and Agriculture, United States Department of Agriculture, Hatch project 7005101 (to C.T.H.). C.T.H. is an H. I. Romnes Faculty Fellow, supported by the Office of the Vice Chancellor for Research and Graduate Education with funding from the Wisconsin Alumni Research Foundation. Research in A.R.鈥檚 lab is also supported by the National Institutes of Health/National Institute of Allergy and Infectious Diseases (R01 AI153356).