Methodological Determinants of Fungal Identification by Proteomic Mass Spectrometry: A Narrative Review with Emphasis on MALDI-TOF MS

1. Introduction

Accurate, rapid identification of fungi is a rate-limiting step in both clinical and environmental mycology. Conventional identification relies on colony morphology, microscopy and biochemical assimilation profiles - methods that are slow, require specialized expertise, and frequently fail to resolve closely related species. With increasing availability of sequence-based techniques, it is well known that microscopy does not predict phylogenetic proximity. Proteomic identification by MALDI-TOF MS addresses these limitations by generating a whole-cell protein mass spectrum, dominated by highly abundant and evolutionarily conserved ribosomal and other housekeeping proteins, and matching that spectrum against a curated reference database. The approach is established for yeasts and is expanding into filamentous fungi (molds), where it still lags behind bacterial and yeast applications in accuracy and standardization.

In yeasts, comparative evaluations have established MALDI-TOF MS to be accurate relative to molecular reference standards for many species, whether measured against conventional phenotypic identification [1,3] or against other MALDI-TOF systems [2].

The apparent simplicity of the MALDI-TOF workflow (smear, extract, spot, match) obscures numerous variables that can shift a spectrum’s quality, the resulting match score, and ultimately whether a laboratory reports a confident species-level identification, a genus-level call, or no identification at all. This review synthesizes these variables within a coherent framework and examines areas where studies differ in the direction or magnitude of reported effects.

The review focuses on MALDI-TOF MS, the dominant proteomic identification platform in current practice, and briefly considers whether these factors extend to complementary approaches such as LC-MS/MS-based workflows.

This is a narrative rather than a systematic review. Sources were identified through targeted searches of PubMed/Europe PMC and Crossref for primary studies and reviews published between 2010 and 2026, combining terms for the platform (MALDI-TOF, mass spectrometry), the organism group (fungi, filamentous fungi, mold, yeast, dermatophytes and dematiaceous fungi), and the variable of interest (identification, spectral library, extraction, colony age, pigmentation, score threshold). Studies were included when they reported identification performance against a molecular or reference standard, or when they reported a directly measured effect of a named variable on spectral quality. No formal risk-of-bias assessment or meta-analysis was performed. Identification rates were not pooled because reference standards, databases, score thresholds, and organism panels differed substantially; this heterogeneity is itself a focus of the review.

2. Biological and Pre-Analytical Determinants

2.1. Colony Age and Growth Phase

Colony age is one of the most contested variables in the fungal MALDI-TOF literature, and the disagreement is instructive.

A large multi-instrument, multi-database comparison of uncommon yeasts, Aspergillus species and rare molds found that identification results obtained after 3 days of incubation were significantly better than those obtained after 8 days, with the magnitude of the effect varying by genus. The authors attributed this finding to changes in protein expression across the growth cycle [6]. This is compounded by a design feature of commercial libraries: reference spectra from both major manufacturers are constructed predominantly from young colonies, so aged colonies lie further from the reference by construction regardless of any change in underlying biology.

In contrast, a mold-database study tested colony ages from 2 to 7 days across four culture media and incubation temperatures from 27 to 42 °C. It found no significant effect of colony age, medium, or temperature on identification success when a standardized extraction protocol was used, and reported 96.5% concordance between duplicate spots [7]. A comparative dermatophyte study likewise notes that while some investigators report that fungal physiology and protein expression are meaningfully affected by culture conditions, others report that identification of at least some organism groups is unaffected by those same conditions [10].

The most defensible interpretation is that the effect of colony age is not a fixed biological constant but may reflect interactions among fungal genus, extraction protocol, and the degree to which local culture conditions resemble those used to construct the reference database. A validated, genus-appropriate extraction protocol may reduce some age-related effects, although this requires confirmation for each organism group and workflow. Colony age should therefore be defined within a locally validated incubation window and specified in the standard operating procedure (SOP), rather than determined solely by apparent adequacy of growth. Inconsistent harvesting may contribute to low identification scores, misidentifications, or failure to obtain an identification.

2.2. Growth Phase, Morphotype and Dimorphism

Beyond chronological age, morphotype is itself a major spectral variable. Thermally dimorphic fungi such as *Histoplasma*, *Blastomyces*, and *Talaromyces marneffei* (formerly *Penicillium marneffei*) grow in a mold phase at ambient temperature (approximately 25 °C) and a yeast or yeast-like phase at body temperature (approximately 37 °C), and each morphotype produces a distinct protein profile. In a study of T. marneffei, the combined Bruker database including the Filamentous Fungi Library failed to identify any of 60 mold- and yeast-phase isolates correctly, misidentifying them as Penicillium funiculosum and P. purpurogenum with subthreshold scores. Only after the database was expanded with spectra from both morphotypes of T. marneffei itself were the remaining isolates correctly identified to species level with scores above 2.0 [11]. The general principle follows directly: a database built from one morphotype cannot be assumed to identify the same species in a different morphotype, and dimorphic fungi require morphotype-specific reference entries and their own validation.

Sporulation is another important consideration, as several studies report that protein patterns change measurably as filamentous fungi progress from vegetative hyphal growth into conidiation, and that this shift is part of why growth-phase standardization matters for reproducible spectra [6]. This is of greater relevance for molds than for yeasts, and it is one reason a protocol validated for yeasts cannot simply be transferred to a mold.

2.3. Culture Medium and Incubation Temperature

Evidence regarding culture medium and incubation temperature is mixed. One well-controlled mold-database study found no effect of medium (Sabouraud dextrose, chromogenic Candida agar, buffered charcoal yeast extract, or blood-containing media) or of incubation temperature from 27–42 °C once extraction was standardized [7]. Other reviews nevertheless continue to list medium and temperature among the factors influencing filamentous fungi identification, alongside colony age and matrix type [6,15,30].

These observations suggest that medium and temperature effects are real but secondary to extraction-protocol quality and that they are most likely to matter for organisms whose secondary metabolism - including pigment production (Section 2.4) - is itself medium-dependent.

2.4. Pigmentation and Melanization

Melanized (dematiaceous) fungi are one of the most mechanistically well-characterized sources of identification failure. Dark fungal pigments absorb and dissipate the laser energy that would otherwise be transferred from matrix to analyte, physically inhibiting desorption/ionization. A foundational study exposing synthetic melanin and Aspergillus niger-derived pigment to MALDI-TOF and MALDI-qTOF analysis found clear signal suppression above pigment concentrations of 250 ng/µL, microscopic examination showed that spectra acquired from visibly darker regions of a deposit yielded poor or absent ion signal while non-pigmented regions of the same deposit ionized normally; adding tricyclazole, a melanin-synthesis inhibitor, to the growth medium suppressed pigmentation and restored analysable spectra [14].

This finding has been replicated and extended. A study differentiating the dematiaceous species Stachybotrys chartarum and S. chlorohalonata obtained good-quality spectra from both hyaline and darkly pigmented conidia only by actively circumventing melanin signal suppression, after which the method distinguished not only the two species but different chemotypes within S. chartarum [13]. A routine-laboratory evaluation of clinical filamentous fungi likewise reported that melanized fungi produce weak spectral profiles with low intensity and few peaks, compounding the separate problem of limited database coverage for these genera - though the same study found that optimized sample preparation and adequate database representation could still achieve high species-level identification rates for dematiaceous genera including Alternaria, Aureobasidium, Cladosporium, Curvularia and Phoma [12]. A review of filamentous fungi identification summarizes the two practical mitigations as growth in liquid media that suppress pigment formation, and pre-analytical washing steps to physically remove excess pigment before matrix application [15].

This distinction is important when interpreting the literature. Melanin-associated signal suppression is a biological influence on the measurement rather than simply an artefact of imprecise technique. However, the extent to which pigmentation impairs identification may depend on the organism, sample preparation, and acquisition conditions. Targeted approaches to reduce pigment formation or remove interfering material may improve spectral quality in selected organisms.

2.5. Taxonomic Complexity: Cryptic Species Complexes and Intraspecific Variation

A substantial share of reported MALDI-TOF misidentifications are not method failures at all but exposures of taxonomic complexity that conventional phenotypic methods could not see. Commonly cited examples include the Candida glabrata species complex (C. glabrata, C. nivariensis, C. bracarensis), the C. parapsilosis species complex (C. parapsilosis, C. metapsilosis, C. orthopsilosis), and the C. haemulonii complex, which includes Candidozyma auris (formerly Candida auris) [18,21].

Similar findings with filamentous fungi have been reported. A large multi-instrument comparison of Aspergillus species and rare molds found that misidentifications concentrated among cryptic species [6]. A study targeting Aspergillus flavus found that while the commercial Bruker score database correctly assigned all 200 isolates to genus level, only 18 of 176 species-level calls (10.2%) were confirmed as true A. flavus, with the remainder called A. oryzae or an unresolved A. flavus/A. oryzae complex - concluding that the commercial database cannot separate these two closely related species, a limitation overcome only by building a dedicated in-house library [22]. Comparable species-complex challenges have also been documented outside clinical mycology. In a spectral database for food-spoilage fungi, a newly described Cladosporium species could only be resolved to a low-discrimination complex-level call because it was absent from the reference database [19]. In a proof-of-concept study of the insect-pathogenic *Metarhizium anisopliae* complex, expansion of the reference library from nine to 11 taxa was required to achieve near-perfect concordance with DNA-based species assignments [20].

The interpretative point is that cryptic diversity converts an apparently method-level problem into a taxonomic one, and the remedy is not better spectra but broader reference coverage together with a molecular standard against which species boundaries can be defined.

3. Sample Processing and Extraction Variables

3.1. Protein Extraction Protocol

Extraction chemistry is an important controllable variable in fungal MALDI-TOF MS, and its effects may vary with organism and cell-wall characteristics. Yeasts, which have comparatively thin and easily lysed walls, often perform adequately with a direct colony smear. Filamentous fungi, whose walls are built from chitin and glucan polymers, generally require those polymers to be disrupted before cytoplasmic and ribosomal proteins can be liberated for ionization. Extraction protocol is therefore decisive in whether an analysable spectrum is obtained at all [6,30].

The quantitative evidence for this is substantial. In a study comparing three sample preparation methods against a Filamentous Fungi Library, species-level identification increased from 48.9% with the simplest preparation to 76.1% with the most laborious full extraction protocol. Validation of a laboratory-specific, lowered score threshold pushed the achievable species-level rate to 86.4% without introducing genus-level misidentifications [29]. However, a more recent prospective comparison of an extraction-free approach against two rapid extraction methods (an NIH protocol and a CDC protocol) across routine and specialized fungal media found, contrary to the expectation that more extensive extraction is invariably superior, that the extraction-free approach outperformed both rapid extraction methods. It yielded correct identification in 90.5% of isolates versus 61.9% and 66.7% respectively, and the gap widened on specialized fungal media (92.9% versus 81.0% and 52.4%) [16].

These findings indicate that the decisive variable is not the stringency of the extraction but probably its alignment with the conditions under which the reference library was constructed, combined with the absence of interfering chemistry. A more intensive protocol may outperform a simpler one when it better matches library-construction conditions, while an extraction-free method may perform well when it avoids interfering chemistry. The evidence does not support the assumption that extraction intensity and accuracy necessarily move together. The relationship depends on the specific chemistry, the organism group and the culture medium, and it should be established empirically in each laboratory rather than inherited from a protocol designed for bacteria [16,29].

3.2. Deposition Technique and Automation

Independent of extraction chemistry, the physical transfer of sample to the target plate is itself a significant source of variability. A validation study compared manual toothpick smearing, semi-extraction, wet formic-acid deposition and a fully automated deposition robot (MALDI Colonyst) across Gram-negative bacteria, Gram-positive bacteria, anaerobes and yeasts. Manual toothpick spotting produced significantly lower and more variable identification scores than every other method tested. When the workflow was scaled to 542 clinical isolates processed under routine conditions, automated deposition again yielded significantly higher scores and lower error rates across all microbial groups, including yeasts [8]. These findings suggest that deposition technique can contribute to score variability, although the magnitude of this effect in filamentous fungi requires direct evaluation.

These findings have potential implications for mold workflows, in which deposits may be heterogeneous. However, because the cited study did not evaluate filamentous fungi, the size of this effect in mold identification remains uncertain. Where automation is unavailable, a standardized manual deposition protocol and operator competency assessment are reasonable quality-assurance measures.

3.3. Matrix Chemistry and Crystallization Homogeneity

Alpha-cyano-4-hydroxycinnamic acid (HCCA) is the near-universal matrix for whole-cell microbial MALDI-TOF identification, selected for its sensitivity across the 0.7–20 kDa mass range in which most diagnostic ribosomal-protein peaks fall [27,28]. Even within a fixed matrix choice, crystallization is rarely homogeneous across a target spot: regions of high analyte or pigment concentration can appear macroscopically different from the surrounding matrix film, and heavily pigmented regions can produce poor or absent ionization even when adjacent non-pigmented regions of the same spot ionize normally [14]. This heterogeneity is one reason acquisition strategies that sample multiple points per spot (Section 4.2) outperform single-point acquisition for recalcitrant fungal targets.

4. Instrument Acquisition Parameters

4.1. Laser Intensity and Shot Number

Fungal whole-cell spectra are intrinsically weaker than bacterial spectra: the thicker, polymer-rich fungal cell wall releases fewer and less abundant proteins into the matrix than a bacterial envelope does, producing spectra with fewer and less intense peaks overall [6]. Because commercial instrument defaults were originally engineered around bacterial signal strength, acquisition parameters tuned for bacteria can leave genuine fungal peaks below the default detection threshold. Multiple reviews therefore recommend higher laser intensity and a greater number of laser shots per spot for fungal targets [6,15].

This has been demonstrated directly. An instrument-optimization study describing a revised acquisition protocol (“Peak MALDI”) found that increasing laser shot number from a default of 240 into a range of 2,000-10,000, paired with a lowered minimum signal-intensity threshold, increased peak number, improved signal-to-noise ratio and increased identification confidence, with the benefit most pronounced for relatively recalcitrant organisms including fungi [9]. Notably, an earlier attempt to improve fungal identification by lowering the minimum signal-intensity threshold alone, without also increasing shot count, produced inconsistent results, because genuine low-abundance diagnostic peaks became harder rather than easier to distinguish from baseline noise [9]. Threshold and shot count are therefore not independent levers, and changing one without the other can degrade performance.

4.2. Signal Threshold, Raster Pattern and Spot Heterogeneity

Randomizing the laser raster pattern across a target spot, rather than repeatedly firing at a single fixed point, compensates for the heterogeneous matrix and analyte distribution described in Section 3.3 by averaging out spot-to-spot and within-spot variability to which single-point acquisition is fully exposed [9]. Multi-spot acquisition and raster randomization address the same underlying problem by different means, and either is preferable to single-point acquisition for molds.

4.3. Calibration and instrument drift

Instrument calibration drifts over the course of a run session, independently of any property of the biological sample, and internal quality control - including automatic calibration against a manufacturer-specified standard before use - is a required rather than optional step in any MALDI-TOF identification workflow [28]. Structural anomalies in the target plate itself, such as scratches or dents, are flagged in the same guidance as a routine troubleshooting checkpoint when spectra are unexpectedly poor [28].

4.4. Platform and Database Cross-Compatibility

Different manufacturers use different scoring conventions, default databases, and, in some cases, different mass spectrometers (including Bruker Microflex/Biotyper, bioMérieux VITEK MS and VITEK MS PRIME, and Autobio Autof MS 1000). A large comparative study tested the same isolates across VITEK MS, VITEK MS PRIME, and Microflex instruments against four databases: two in vitro diagnostic databases, one research-use-only Bruker filamentous fungi library, and the independent academic MSI-2 database. It found meaningful differences not only in overall identification rate but in the rate of spectra classified as simply non-analysable: 15% for VITEK MS PRIME versus 3% for the older VITEK MS platform in the same study. The same study found that the academic MSI-2 database outperformed every commercial database for rare and cryptic organisms, at the cost of being less standardized for routine use [6]. Platform choice therefore interacts with database choice in ways that are not predictable from either factor alone, and newer instrument hardware does not automatically translate into better identification performance if the paired database or acquisition defaults are not equally mature. Cross-platform transfer of thresholds or protocols requires independent validation; nothing in this literature supports assuming equivalence.

5. Database and Bioinformatic Factors

5.1. Reference Library Completeness

Across essentially every study reviewed here that compared a commercial-only database against an expanded or in-house database, library completeness emerged as the single largest lever available to a laboratory. A large filamentous fungi study testing ten identification algorithms across 422 isolates spanning 126 species found that an in-house database achieved 87.41% overall identification versus 35.15% for the Bruker commercial database alone; restricting the analysis to species with at least three reference strains improved the in-house figure to 92.1% versus 46.6% [4]. An independent in-house library validation study in Medellin, Colombia found a closely comparable pattern: 88.70% species-level identification with an expanded in-house library versus 44.44% with the commercial library alone, with most of the remaining gap attributable to two genera, Neoscytalidium and Sporothrix, that were essentially absent from the commercial spectral database [5]. The T. marneffei case described in Section 2.2, in which the commercial database misidentified every isolate until morphotype-specific reference spectra were added [11], is a further instance of the same phenomenon.

These studies support the importance of library completeness, although the magnitude of benefit may vary by laboratory, organism panel, platform, and database. Including relevant reference strains, with appropriate representation of within-species diversity, can improve the likelihood of identifying organisms that are poorly represented in commercial libraries; the relationship should not be assumed to be directly proportional.

5.2. Score-Threshold Interpretation

Score thresholds validated for bacterial identification - historically a log score of 2.0 for species-level confidence and 1.7 for genus-level confidence on the Bruker platform, are not necessarily appropriate for fungi, particularly molds and dermatophytes. Multiple laboratories have found that a lowered, locally validated threshold, commonly around 1.7 rather than 2.0, can reduce the number of unreported isolates without introducing new misidentifications, provided the threshold change is itself validated against a reference method [4,9].

Two caveats apply. First, thresholds are platform-specific: platforms report scores on different underlying scales (for example, some Autobio and VITEK systems use a 0-10 or percentage-style confidence scale rather than the Bruker log-score system), so a threshold calibrated on one platform cannot be transferred to another without independent validation [6]. Second, a lowered threshold is defensible only in conjunction with the acquisition changes described in Section 4.1: lowering a threshold without increasing shot count risks admitting noise rather than recovering signal [9].

5.3. Software, Algorithm and Version Considerations

Beyond the reference spectra themselves, the software that compares an acquired spectrum against the database is an additional variable. Supplementary statistical classification tools such as ClinProTools have resolved identification problems, for example within the C. haemulonii complex, that the standard Biotyper and VITEK MS pattern-matching algorithms could not resolve using their default comparison methods [23]. As acquisition software, peak-picking algorithms and manufacturer database versions are updated over time, comparative performance figures generated on one software version may not generalize to newer releases. The comparative literature synthesized here should therefore be read as a snapshot of reported performance at a point in time rather than a permanent ranking, and laboratories should record the software and database versions under which their local validation was performed.

6. Specimen-Type and Clinical-Context Factors

Most factors discussed so far apply to identification from a pure cultured isolate. A distinct set of challenges arises when proteomic identification is attempted directly from a clinical specimen, such as positive blood culture broth, cerebrospinal fluid or other body fluids bypassing subculture. It should be noted at the outset that CLSI M58, the primary consensus framework for MALDI-TOF identification, explicitly excludes direct-from-specimen identification from its scope [28]; guidance for that workflow therefore rests on the primary literature rather than on a consensus standard.

A review of direct MALDI-TOF identification from bloodstream infections and sterile body fluids found that, although direct identification can deliver results within 30–60 minutes, success is not uniform across organism classes: Gram-negative bacteria are identified more successfully than Gram-positive bacteria, which in turn are identified more successfully than fungi, a gradient attributed to differences in cell-envelope structure and protein-extraction efficiency [24]. The same review identified four recurring constraints: limited sensitivity in samples with low organism burden, spectral interference from host-derived proteins, difficulty resolving polymicrobial infections in which more than one organism contributes overlapping peaks to a single spectrum, and continuing gaps in reference databases for emerging pathogens [24].

Host-matrix interference is a mechanistically well-described problem. Non-microbial proteins present in blood culture broth or whole blood disturb the proteome profile during direct identification, which has motivated dedicated preprocessing steps. In one large validation involving 2,032 positive blood-culture samples, density centrifugation was combined with chemical lysis and extraction to remove interfering host- and broth-derived proteins while concentrating the microbial signal before spotting [25]. A case report of direct whole-blood identification in a paediatric oncology patient provides an illustrative example. The authors noted that host-derived material can greatly exceed microbial material and that haemoglobin, albumin, and anticoagulants may interfere with direct identification. Successful direct-from-whole-blood identification may therefore depend on a sufficiently high organism burden [26]. General reviews of MALDI-TOF workflow design likewise note that body fluid specimens typically require dedicated purification steps before the standard identification pipeline can be applied [27].

Prior antifungal exposure is a possible but insufficiently studied source of variability. MBT-ASTRA and related rapid susceptibility assays detect growth-associated spectral changes under drug exposure; however, this does not establish that prior antifungal treatment alters routine identification spectra. The potential interaction should be considered a hypothesis requiring direct study, rather than an established determinant of identification performance. Reference 31 should be replaced or supplemented if a direct primary study supporting this specific claim is identified.

7. Beyond MALDI-TOF: Considerations for Other Proteomic Approaches

The same biological variables—including colony age, growth phase, pigmentation, and taxonomic complexity—may affect other proteomic approaches applied to fungal cells or lysates, including LC-MS/MS-based shotgun proteomics and targeted analyses used for biomarker discovery and resistance-mechanism profiling. LC-MS/MS introduces additional sources of variability, including protein-digestion efficiency, chromatographic reproducibility, ionization efficiency, and instrument performance. Its greater analytical depth may help distinguish organisms that are difficult to resolve using whole-cell MALDI-TOF MS fingerprinting, although performance depends on the analytical workflow and reference data available.

This has implications for the interpretation of the literature reviewed here. Identification rates quoted for whole-cell MALDI-TOF reflect the specific combination of reference library, protocol, and threshold employed; they do not represent an upper bound on what proteomic identification of fungi can achieve. A full treatment of LC MS/MS specific variability lies beyond the scope of this review, which is centred on the whole-cell fingerprinting workflow used in routine and reference laboratory practice.

8. Summary

Three higher-order patterns emerge from the preceding sections.

First, variability attributed to fungal biology may be confounded by methodological variability. For colony age, one study using a standardized extraction protocol found no significant effect across the tested age range [7], whereas another reported better identification from younger colonies [6]. These studies differ in multiple respects, so the role of extraction standardization cannot be isolated with certainty. The same pattern recurs for culture medium and temperature. The effects of dimorphism and melanization, however, are supported by mechanistic evidence and represent genuine sources of spectral variation (Section 2.2 and Section 2.4). Variability should therefore be interpreted in the context of the protocol and organism studied rather than as a fixed property of the organism.

Second, two particularly actionable determinants are reference-library completeness and standardized sample preparation and deposition. The commercial-versus-expanded database comparisons reviewed here generally favoured the expanded libraries [4,5,11]. A deposition study also found advantages for standardized or automated methods over manual techniques, although it evaluated bacteria and yeasts rather than filamentous fungi [8]. These two variables are also the most actionable at the level of an individual laboratory, in contrast to instrument hardware or manufacturer database architecture, which are largely fixed once a platform is purchased. Table 1 summarizes the direction, magnitude and strength of evidence for each factor; the ordering of that table is itself the practical message, with the strongest and most actionable evidence at the top.

Third, the field’s central unresolved challenge is not accuracy on common pathogens but accuracy on the tail of rare, cryptic, melanized, dimorphic and newly emerging organisms - precisely the isolates most likely to be ambiguous and most likely to require a correct identification. Commercial systems, now well optimized for frequent pathogens, continue to underperform specifically in this tail, and closing that gap depends on the same two levers - database expansion and extraction protocol validation - applied to underrepresented taxa, rather than on any single instrumental breakthrough [4,5,11,12,22].

Limitations. This is a narrative review, and several limitations follow. The search was targeted rather than systematic, so the studies discussed are not an exhaustive census and negative or unpublished findings are likely underrepresented. Identification rates are not pooled because the underlying reference standards, databases, thresholds and organism panels differ too much for pooling to be meaningful, which means the magnitudes quoted throughout are study-specific. Evidence for direct-from-specimen workflows is derived largely from specialized workflows and studies not designed specifically for fungal identification. Biosafety inactivation evidence includes direct fungal data as well as related evidence from mycobacterial systems; these should be distinguished when interpreting the strength of evidence. Finally, performance figures are tied to the software and database versions in use when the studies were performed (Section 5.3) and should be treated as a snapshot rather than a permanent ranking.

9. Practical Implications for Laboratory Quality Assurance

CLSI M58 provides guidance for implementation of MALDI-TOF MS identification of cultured microorganisms, including isolate preparation, quality control, method verification, competence assessment, troubleshooting, and result interpretation and reporting [28]. Interpreted in light of the evidence synthesized above, several specific practices are warranted beyond the framework’s general guidance. These are collected in Box 1 and are ordered by the strength of the evidence behind them.

Two practices merit particular emphasis. First, extraction intensity should not be assumed to correlate directly with accuracy; extraction chemistry should be selected and validated for the relevant organism groups and workflow rather than chosen solely for its intensity [16,29]. Second, any lowered score threshold should be evaluated within a locally validated workflow. If acquisition settings are also changed, their combined effects should be assessed against an appropriate reference method to ensure that improved identification yield does not compromise accuracy [9].

Table 2. Direct comparisons of extraction and deposition protocols in fungal MALDI-TOF MS.

Study	Methods compared	Organisms	Result
Honsig et al., 2022 [29]	Three sample preparation methods against a filamentous fungi library, plus locally validated score threshold	Clinical filamentous fungi	Species-level identification 48.9% to 76.1% with the most thorough protocol; 86.4% after threshold validation, without new genus-level misidentifications
Olaniyan et al., 2026 [16]	Extraction-free versus two rapid extraction methods (NIH and CDC protocols), on routine and specialized fungal media	Filamentous fungi	Extraction-free 90.5% correct versus 61.9% and 66.7%; on specialized media 92.9% versus 81.0% and 52.4%
Lau et al., 2013 [7]	Simple procedure on solid media; four media; 27-42 C; colony ages 2-7 days	Molds	No significant effect of age, medium or temperature once extraction was standardized; 96.5% concordance between duplicate spots
Chudejova et al., 2017 [8]	Manual toothpick, semi-extraction, wet formic acid, automated deposition robot	Gram-negative, Gram-positive, anaerobes, yeasts	Manual toothpick worst and most variable; automated deposition gave higher scores and lower error rates at routine scale

Table 2. Direct comparisons of extraction and deposition protocols in fungal MALDI-TOF MS.

Study	Methods compared	Organisms	Result
Honsig et al., 2022 [29]	Three sample preparation methods against a filamentous fungi library, plus locally validated score threshold	Clinical filamentous fungi	Species-level identification 48.9% to 76.1% with the most thorough protocol; 86.4% after threshold validation, without new genus-level misidentifications
Olaniyan et al., 2026 [16]	Extraction-free versus two rapid extraction methods (NIH and CDC protocols), on routine and specialized fungal media	Filamentous fungi	Extraction-free 90.5% correct versus 61.9% and 66.7%; on specialized media 92.9% versus 81.0% and 52.4%
Lau et al., 2013 [7]	Simple procedure on solid media; four media; 27-42 C; colony ages 2-7 days	Molds	No significant effect of age, medium or temperature once extraction was standardized; 96.5% concordance between duplicate spots
Chudejova et al., 2017 [8]	Manual toothpick, semi-extraction, wet formic acid, automated deposition robot	Gram-negative, Gram-positive, anaerobes, yeasts	Manual toothpick worst and most variable; automated deposition gave higher scores and lower error rates at routine scale

Box 1. Evidence-grounded quality-assurance practices

• Fix colony age within a validated, young incubation window matched as closely as possible to the age used to construct the reference database, and state that window in the standard operating procedure rather than harvesting on the basis of apparent adequacy of growth.
• Treat dimorphic fungi as requiring morphotype-specific reference spectra and independent validation, rather than a single species-level entry.
• For genera known to be melanized, validate a pigment-suppression or pigment-removal step (medium-based suppression or a pre-analytical wash) rather than assuming default extraction will succeed.
• Select and validate extraction chemistry per organism class (yeast, mold, dimorphic fungus) rather than applying a single protocol uniformly, and re-validate empirically rather than assuming that more aggressive extraction is superior.
• Automate deposition where feasible, and where it is not, standardize manual deposition with competence assessment since the effect of spotting technique on score variability is large.
• For fungal targets, increase laser shot count beyond bacterial defaults, and pair any threshold adjustment with a shot-count increase rather than adjusting threshold alone.
• Recalibrate every run session and inspect target-plate integrity as a first troubleshooting step when spectra are unexpectedly poor.
• Where a locally lowered score threshold is adopted, validate it explicitly against a molecular reference method before implementation, and document that validation under the M58 method-verification framework.
• For rare, cryptic or locally prevalent taxa not well represented in commercial databases, invest in in-house library expansion rather than relying on periodic commercial updates alone.
• Record the software, database and reference-library versions under which local validation was performed, so that performance claims remain interpretable after version changes.

10. Conclusion

Identification of fungi by proteomic mass spectrometry is not a single uniform assay but a workflow assembled from a long chain of biological, chemical and instrumental choices, each of which can independently move an isolate from a confident species-level call to an unreliable or absent identification. The literature synthesized here identifies at least a dozen distinct, evidence-supported causes of variability spanning colony biology, sample processing, instrument operation and database architecture. Several of these, notably colony age and culture medium, remain genuinely contested rather than settled.

Database completeness and technique standardization are the two variables most within an individual laboratory’s control and most consistently associated with large improvements in identification accuracy. A practical first step when an organism cannot be identified is to determine whether it is adequately represented in the reference library and whether the local protocol has been validated for that organism group. These considerations should be assessed alongside the possibility that the organism is intrinsically difficult to resolve by the chosen method. Future work would benefit from applying the rigor already used to standardize bacterial MALDI-TOF workflows specifically to the fungal tail of rare, cryptic, dimorphic and melanized organisms, where the current evidence base and current database coverage both remain thinnest.