Fungal identification using conventional methods in routine practice often faces challenges, such as the need for culti-vation, being time-consuming, and low specificity. With the introduction of antifungal stewardship programs and routine antifungal susceptibility testing for managing fungal infections, precise identification of fungal pathogens has become in- creasingly important. Identification using DNA barcoding markers has been recognized as another "gold standard" method for fungal identification; however, this method entails high costs, complex protocols, and long turnaround times. The internal transcribed spacer (ITS) region of ribosomal DNA has been widely adopted as the primary DNA barcode for fungal identification due to several key attributes^1,2, including the availability of comprehensive reference databases with type specimens to discriminate most species, standardized amplification protocols, and a phylogenetic framework that allows evolutionary interpretation. These features collectively establish ITS sequencing as the current reference standard for definitive fungal identification, particularly for species that cannot be reliably distinguished by phenotypic methods. Since MALDI-TOF MS was developed^3,4, many studies have applied the MS to fungal identification, especially for yeast pathogens in routine diagnostics^5-11, to overcome DNA barcoding dis- advantages. MALDI-TOF MS identifies microorganisms by analyzing their protein profiles, generating unique mass spectra patterns that serve as molecular fingerprints for each species. This technique has gained popularity in clinical microbiology due to its rapid turnaround time (minutes vs. days) and low operational costs^12,13.

In the present study, we compared and assessed two molecular methods, MALDI-TOF MS and DNA barcoding using ITS, to identify yeast strains isolated from patients and domestic environments.

1. Collection of yeast strains

One hundred and fifteen yeast strains were collected from 2022 to 2025. Of these 89 strains (strains F1-F54 and B2-B57) were isolated from clinical specimens of patients who were admitted at the Srinagarind Hospital, Khon Kaen, Thailand. The remaining 26 strains (strains E3-E39) were collected from household environments within the Khon Kaen municipal area. The colonial morphology of each strain was examined using Sabouraud dextrose agar and CHROMagar (BrillianceTM Candida Agar, Oxoid, UK), while micromorphology was done using lactophenol cotton blue and/or nigrosin staining.

2. Identification by MALDI-TOF MS

All yeast strains were identified using MALDI-TOF MS (MALDI Biotyper®, autoflex® maX mass, Bruker Daltonics, Bremen, DE). The method was performed according to the manufacturer's protocol. Sample preparation followed the manufacturer's formic acid extraction protocol. Briefly, yeast cells were smeared onto a ground steel MALDI target plate (MTP 384), air-dried, overlaid with 1 μl of 70% formic acid and dried, then overlaid with 1 μl of α-cyano-4-hydroxycinnamic acid (HCCA) matrix solution and dried again. The steel plate was examined to obtain mass spectral profiles using an Autoflex maX mass spectrometer. Each yeast strain profile was identified using Bruker's MALDI Biotyper® Compass soft- ware version 4.1.100 with the MBT 8468 Species/Entry List Revision C (March 2019), and identifications were accepted according to the manufacturer's criteria.

3. PCR sequencing of ITS

All strains were DNA-extracted using a DNA extraction kit (Vivantis Technologies Sdn Bhd., Selangor Darul Ehsan, Malaysia). The DNAs were quantified using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Massachusetts, USA). The ITS sequence in the ribosomal gene was amplified by PCR using the forward primers ITS1 (TCCGTAGGTGAACCTGCGG), V9G (TTACGTCCCTGCCCTTTGTA)¹⁴, or ITS5 (TCCTCCGCT- TATTGATATGC)¹⁵ and the reverse primers including ITS4 (GGAAGTAAAAGTCGTAACAAGG)¹⁵ or LS266 (GCATTCCC- AAACAACTCGACTC)¹⁶. Different primer combinations were used to optimize amplification success across diverse yeast taxa. The primer pair ITS1/ITS4 was primarily used, while alter- native primer pairs (V9G/LS266 or ITS5/ITS4) were employed when amplification failed or produced weak products. The PCR reaction contained 10 μM of each primer, 12.5 μl of KOD One® PCR Master Mix (TOYOBO, JP), and >10 ng genomic fungal DNA in 25 μl of total volume reaction. The PCR conditions started at 95℃ for 5 min, followed by 35 cycles of denaturation at 95℃ for 35 sec, annealing at 48℃ for 30 sec, and extension at 72℃ for 30 sec. Amplification was performed using a Veriti® Thermal Cycler (Applied Biosystems, CA, USA). The amplification products were examined with 2% gel electrophoresis and submitted to Sanger's sequencing.

4. Identification by ITS barcoding

The ITS sequences obtained from Sanger's sequencing were edited using BIOEDIT v.7.2.5¹⁷ and blasted against the type materials in GenBank using megablast in BLAST® blastn¹⁸. The datasets, including the sequences of the tested strains and those retrieved from megablast, were prepared using BIOEDIT v.7.2.5 and aligned with the algorithms, specifically ClustalW in BIOEDIT v.7.2.5, Multiple Alignment using Fast Fourier Transform (MAFFT)¹⁹, or MUltiple Sequence Com- parison by Log-Expectation (MUSCLE)20 via the EMBI-EBI web server²¹. Phylogenetic analyses for species identification were performed using MEGA11²². The best model and appro- priate phylogenetic algorithms were inferred to reconstruct a phylogenetic tree and to estimate percentage similarity. Species identification was considered valid when the test strain formed a monophyletic clade with the type species in the phylogenetic tree and/or showed sequence similarity with the type specimen across the ITS region.

5. Assessment of MALDI-TOF MS for yeast identification

With ITS barcoding serving as the reference method, the identification rate and accuracy of MALDI-TOF MS were cal- culated using the following formulas:

Agreement between the two methods was assessed using Cohen's kappa coefficient, calculated for two scenarios: (1) all isolates and (2) isolates of species represented in the MALDI-TOF MS database. Statistical analysis was performed using SPSS version 29.0 (IBM Corp., Armonk, NY, USA). The Cohen's kappa (κ) coefficients were interpreted based on the criteria established by Landis and Koch²³: ≤ 0 (no agreement), 0.01-0.20 (slight), 0.21-0.40 (fair), 0.41-0.60 (moderate), 0.61-0.80 (substantial), and 0.81-1.00 (almost perfect agree- ment).

1. Identification by MALDI-TOF MS

MALDI-TOF MS identified 86 strains belonging to 11 species (number of strains), including C. albicans (46), C. tropicalis (22), C. metapsilosis (2), C. parapsilosis (1), Meyerozyma guil- liermondii (syn. Candida guilliermondii) (1), Nakaseomyces glabratus (syn. Candida glabrata) (7), Pichia kudriavzevii (syn. Candida krusei) (2), Diutina rugosa (syn. Candida rugosa) (1), Kodamaea ohmeri (1), Cystobasidium minutum (syn. Rhodotorula minuta) (1), and Rhodotorula mucilaginosa (1). However, the MS failed to identify 29 strains. All results are shown in Table 1.

2. Identification by ITS barcoding

All 115 yeast strains were identified using phylogenetic analysis of their ITS sequences in comparison with the type materials. Species identification was determined based on monophyletic clustering and/or the highest sequence similarity with the type species, as shown in Fig. 1. Of the 115 strains, 102 were classified as ascomycetous yeasts belonging to 14 species, whereas 13 strains were classified as basidiomycetous yeasts, comprising 5 species and 1 species complex. The ascomycetous species (number of strains) included Candida albicans (46), C. tropicalis (22), Nakaseomyces glabratus (8), C. orthopsilosis (6), C. metapsilosis (4), Pichia kudriavzevii (4), Wickerhamomyces sydowiorum (3), C. parapsilosis (2), Debaryomyces nepalensis (2), Meyerozyma caribbica (1), Kodamaea ohmeri (1), Diutina mesorugosa (1), Trichomona- scus ciferrii (syn. Candida ciferrii) (1), and Yarrowia lipolytica (1). The basidiomycetous sp ecies (number of strains) included Papiliotrema rajasthanensis (5), Rhodotorula mucilaginosa (3), Kwoniella heveanensis (2), Cystobasidium benthicum (1), Cystobasidium minutum (1), and Naganishia albidosimilis-liquefaciens (syn. Cryptococcus albidosimilis and C. lique- faciens) (1). Candida albicans, C. tropicalis, C. parapsilosis complex, and Nakaseomyces glabratus were the predominant species isolated from patients' specimens, whereas basidio- mycetous species predominated in environmental specimens. All details are presented in Table 1.

Figure 1. Phylogenetic tree of 115 yeast strains and 41 type materials according to a dataset of 156 ITS sequences. The dataset was aligned to 951 positions by MAFFT and analyzed in MEGA11 by the Neighbor-joining method with the Tamura 3-parameter + G model having 1,000 bootstrap replications. Number on the nodes = % bootstrap support, ≥50% nodes shown. Number behind identified species = % similarity to the type-material

Modify Delete

3. Assessment of MALDI-TOF MS for yeast identification

MALDI-TOF MS and ITS barcoding results are compared in Table 1. In addition to the 29 unidentified strains, the MS misidentified three strains as Candida parapsilosis, Diutina rugosa, and Meyerozyma guilliermondii; these were sub- sequently identified as C. orthopsilosis, D. mesorugosa, and M. caribbica, respectively, based on ITS sequencing. Using ITS barcoding as the reference method, the overall identification rate for MALDI-TOF MS was 74.78%, with an overall accuracy of 96.51%. The MS system demonstrated the highest identi- fication rate for Candida species (100%), whereas the lowest rate was observed for basidiomycetous species (15.38%). Specifically, accuracy values were 98.61% for Candida species, 83.33% for non-Candida species, and 100% for basidio- mycetous species. Cohen's kappa analysis demonstrated substantial agreement between ITS barcoding and MALDI-TOF MS across all isolates (n = 115, κ = 0.65, p < 0.001). This agreement strengthened when the analysis was restricted to isolates of species represented in the MALDI-TOF MS database (n = 100, κ = 0.76, p < 0.001). Moreover, almost perfect agreement (κ = 0.81, p < 0.001) was achieved for common pathogenic Candida species, including C. albicans, C. tropicalis, and the C. parapsilosis complex. Detailed results are shown in Table 2.

Invasive fungal infections, particularly candidemia and in- vasive aspergillosis, are associated with mortality rates of 20-40% and 30-95%, respectively, depending on the patient population and timeliness of appropriate therapy^24,25. Early identification has been shown to improve survival outcomes²⁶. Recent stewardship studies have demonstrated that rapid diagnosis can reduce inappropriate antifungal use and shorten hospital stays²⁷.

In our study, DNA barcoding using ITS yielded satisfactory results, confirming its role as another "gold standard" method for yeast identification. The well-supported phylogenetic species clusters and/or the minimal genetic distances to type strains serve as the criteria for sequence-based identification, as previously mentioned^28,29. DNA barcoding represents a highly robust methodology for species identification, as its procedures align with the foundational protocols established in contemporary yeast taxonomy through molecular phylo- genetic analysis. However, to ensure precise identification, DNA barcoding necessitates the integration of molecular techniques and deep taxonomic knowledge. While BLAST® blastn¹⁸ appears to be an efficient tool for identifying DNA barcoding sequences, results can sometimes be difficult to interpret without sufficient expertise. Although modern fungal taxonomy relies on the genealogical species concept, challenges persist regarding taxonomic non-consensus and species complexes. Assigning a single barcoding marker often proves problematic for species defined via multigene analysis or those with ambiguous classifications. Therefore, phylo- genetic analysis, including comparisons with type materials, is essential to achieve accurate identification. Our study utilized a straightforward phylogenetic approach using ITS, a method rarely incorporated into MALDI-TOF assessment studies. This approach successfully identified nearly all patho- genic species at both the species and species complex levels. Notably, ITS cannot differentiate two basidiomycetous species, Naganishia albidosimilis and N. liquefaciens (syn. Cryptococcus albidosimilis and C. liquefaciens), due to their identical ITS sequences. However, these species can be distinguished through phylogenetic analysis of an alternative marker, the D1/D2 domain of the LSU rDNA³⁰. Despite its high resolution, DNA barcoding remains limited by long turnaround times and high operational costs, which are significant drawbacks for routine clinical use.

Across all species investigated, MALDI-TOF MS showed sub- stantial agreement with ITS barcoding (κ = 0.65), with an identification rate of 74.78% and a high accuracy of 96.51%. However, the system failed to identify seven species not represented in its database, highlighting a primary limitation of MALDI-TOF MS. These species included Meyerozyma caribbica, Debaryomyces nepalensis, Wickerhamomyces sydo- wiorum, Diutina mesorugosa, Cystobasidium benthicum, Kwoniella heveanensis, and Papiliotrema rajasthanensis. When the analysis was restricted to species represented in the database, the MS demonstrated higher agreement with ITS barcoding (κ = 0.76), with the identification rate and accuracy reaching 84% and 98.91%, respectively. Identification rates for uncommon yeasts and molds typically range from 65% to 85%, depending on the instrument and database¹⁰. Although the MS database used in this study included many species (Table 2), it still failed to identify certain strains. This may be attributed to factors such as the partial extraction method used for the specimens³¹ or strain variations in mass spectral profiles³².

The overall identification accuracy in our study was highly satisfactory at 96.51%, reaching 98.81% when restricted to species present in the database. Such high accuracy (often > 95%) is consistent with findings for common yeast identification using either the Bruker MALDI Biotyper or the bioMérieux VITEK MS/MS^33-35. MALDI-TOF MS, when com- bined with curated libraries or purpose-built datasets, has been shown to successfully differentiate species complexes, including C. parapsilosis, C. guilliermondii, C. glabrata, and Cryptococcus neoformans/gattii ^8-11.

However, in this study, the MS failed to differentiate the species within the C. parapsilosis complex and the Meyero- zyma guilliermondii (syn. Candida guilliermondii) complex. In addition to technical limitations, incomplete databases, and variations in mass spectral profiles, the close genetic relation- ships among these species could be a plausible explanation for identification failure. MALDI-TOF MS identifies organisms based on ribosomal protein mass spectra, which are highly conserved among closely related species³⁶. When species share more than 98% sequence similarity, discriminatory protein peaks may be subtle or absent, resulting in misidentification or non-identification. This phenomenon has been documented in other fungal complexes, including Aspergillus niger, A. flavus, and the A. nidulans complex³⁷, as well as the Crypto- coccus gattii complex¹¹. Similarly, the close phylogenetic relationship between C. albicans and C. africana poses signifi- cant challenges for MALDI-TOF MS³⁸. Database expansion alone may not fully resolve this issue. Species complexes with minimal protein-level divergence may require supplementary identification methods, such as ITS sequencing or multilocus sequence typing, for definitive differentiation³⁹.

From a clinical perspective, these misidentifications may have limited therapeutic impact. Species within the C. para- psilosis complex generally exhibit similar antifungal suscepti- bility profiles, with minimal differences in minimum inhibitory concentration distributions for azoles and echinocandins^40,41. Similarly, M. caribbica and M. guilliermondii show comparable susceptibility patterns⁴². No inter-complex misidentifications that could lead to inappropriate antifungal selection were observed in our cohort. However, accurate species-level iden- tification within complexes remains important for epidemio- logical tracking and understanding virulence differences. For instance, Candida orthopsilosis and C. metapsilosis differ from C. parapsilosis in biofilm formation capacity and tissue tropism⁴³. Such information is valuable for infection control and outbreak investigation, even when immediate therapeutic decisions are unaffected.

In our findings, when focusing on the most common patho- genic Candida species present in the database, MALDI-TOF MS showed almost perfect agreement with ITS barcoding (κ = 0.81), achieving 100% identification rate and accuracy in C. albicans and C. tropicalis. Furthermore, the MS proved highly reliable in identifying Nakaseomyces glabratus (syn. Candida glabrata) with 100% accuracy. To improve the identification capability and accuracy of MALDI-TOF MS, expanding the database to include a broader range of species associated with human and animal infections, such as Meyerozyma caribbica^42,44, Diutina mesorugosa⁴⁵, and Debaryomyces nepalensis⁴⁶, should be considered a critical priority.

Additionally, our findings revealed that members of the C. parapsilosis complex, viz., C. parapsilosis, C. orthopsilosis, and C. metapsilosis, were isolated from various environmental surfaces, including floors, electric fans, air conditioners, air purifiers, trash bins, and refrigerators. These observations are consistent with their established environmental persistence and hand-mediated transmission pathways⁴⁷. These results highlight the critical role of environmental surveillance in monitoring the dissemination of azole-resistant strains⁴⁸ or potential sources of infection⁴⁹.

In conclusion, both MALDI-TOF MS and ITS barcoding have distinct roles in clinical mycology laboratories. For cost-effective laboratory practice, we recommend MALDI-TOF MS as the first-line method for rapid identification of com- mon yeast pathogens. ITS barcoding should be reserved for unidentified isolates, species complexes, rare species, and high-impact clinical cases requiring definitive identification^39,50-52. Continued expansion of MALDI-TOF reference databases, particularly to include closely related or cryptic species as well as newly described species or discovered pathogens, will further enhance its utility.