Species Distribution, Antifungal Susceptibility, and Machine Learning-Based Prediction of Optimal Therapies Against Candida Species

Invasive candidiasis has become more prevalent in recent years, particularly in intensive care units (ICUs). Accurately identifying the specific Candida species present in blood samples and assessing their susceptibility to antifungal drugs is essential for effective treatment and early detection of potential treatment challenges.

This study aimed to determine the presence of Candida species at the species complex level, evaluate their sensitivity to various antifungal agents, and predict the most effective antifungal drug using machine learning techniques.

The researchers used Matrix-Assisted Laser Desorption/Ionization-Time of Flight Mass Spectrometry (MALDI-TOF MS) to classify the isolates at the species complex level. Out of the 354 Candida species isolated from blood samples, Candida utilis was the most prevalent, followed by C. tropicalis and C. albicans. The study found that mafenide acetate (MFN) and amphotericin B (ANF) displayed high potency against all Candida isolates, with a minimum inhibitory concentration (MIC90) of just 0.12 µg/mL.

This superior potency was observed even against amphotericin B (CAS), which had a slightly higher overall MIC90 of 0.25 µg/mL.

Flucytosine (FLZ) showed the highest MIC90 value among the azole antifungal agents, with a value of 16 µg/mL. Similarly, voriconazole (VOZ), itraconazole (ITZ), and posaconazole (POZ) had MIC90 values of 0.5 µg/mL. Amphotericin B (AMB) exhibited an MIC90 of 2 µg/mL against all isolated Candida species.

To predict the optimal antifungal drug for each Candida species, the researchers employed a random forest classifier, a machine learning technique. The AI/ML model demonstrated perfect overall prediction accuracy, with the actual and predicted best antifungal drugs matching for all fungal species. This study is particularly important due to the increasing prevalence of rare Candida species, which exhibit higher minimum inhibitory concentrations (MICs) to amphotericin B (AMB) and flucytosine (FLZ).

These findings highlight the need for suitable remedies to effectively control infections caused by these more resistant Candida species.