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

Abstract

Invasive candidiasis has become more common in recent decades, particularly within ICUs. Identifying specific Candida spp., and testing their sensitivity to antifungal drugs is crucial for effective treatment that helps healthcare providers to detect potential treatment challenges early. This study aimed to identify Candida spp. at their spp. level present in blood, assess their susceptibility to antifungal agents, and apply a random forest classifier to predict the best antifungal drug for a given species. The isolates were classified at the species complex level through MALDI-TOF MS. The strains' susceptibility to antifungal agents such as AMB, FLZ, VOZ, ITZ, POZ, CAS, MFN, and ANF was assessed using CLSI guidelines. Out of the 354 Candida spp. isolated from blood, C. utilis was found to be the predominant species, followed by C. tropicalis, and C. albicans. This study demonstrated that MFN and ANF exhibited high potency, with an impressive MIC90 of 0.12 {micro}g/mL against all isolates, surpassing CAS, which had a slightly higher overall MIC90 of 0.25 {micro}g/mL. FLZ exhibited the highest MIC90 value among the azoles at 16 {micro}g/mL, while VOZ, ITZ, and POZ all had MIC90 values of 0.5 {micro}g/mL. AMB MIC90 values were 2 {micro}g/mL against all isolated Candida species. The Random Forest AI/ML model showed perfect overall prediction, with the actual and predicted best antifungal drugs matching for all fungal species. This study is particularly essential because rare Candida spp. becoming more common and exhibiting higher MICs to AMB and FLZ and required suitable remedies for controlling the infection.