Weekly Fungal Diagnostics Update
Emerging Diagnostics and Technology Watch
This week’s horizon scan identified four developments from wider diagnostic microbiology that could influence future fungal testing. The strongest themes are improved recovery of microbial DNA from blood, faster sequencing workflows, and the use of artificial intelligence to support—not simply replace—laboratory staff.
None of these studies currently supports an immediate change in routine fungal diagnostic pathways. However, several address technical or operational barriers that are directly relevant to mycology.
1. Same-day sequencing after blood-culture enrichment
Sajib MSI, Oravcová K, Brunker K, et al. Rapid and modular workflows for same-day sequencing-based detection of bloodstream infections and antimicrobial resistance determinants using culture-enriched samples. Microbiology Spectrum. 2026;14(7):e0324025.
Diagnostic development: Rapid metagenomic sequencing following short blood-culture enrichment.
Technology-watch rating: ★★★ High priority
What did the researchers study?
The investigators developed a rapid chemical host-DNA depletion protocol, termed M-15, for use with culture-enriched blood samples. The workflow was assessed with and without Oxford Nanopore adaptive sampling, which can selectively enrich microbial sequence data during sequencing.
The aim was to shorten the time required for pathogen identification and antimicrobial-resistance prediction after collection of blood from patients with suspected bloodstream infection.
What did they find?
Using a selected abundance threshold, M-15 metagenomic next-generation sequencing achieved reported species-identification sensitivity of 93.75% and specificity of 100% in the study dataset.
The workflow could potentially provide species and antimicrobial-resistance information within:
- approximately 5–7 hours after blood-culture positivity; or
- approximately 13–15 hours after specimen collection when rapid culture enrichment was successful.
The researchers compared the method with commercial and previously published host-DNA depletion approaches. Resistance-gene predictions were also compared with available phenotypic susceptibility results.
Strengths
- It addresses an important source of delay in bloodstream-infection diagnosis.
- It combines organism identification with resistance-gene detection.
- The modular workflow could potentially be added to existing blood-culture systems.
- Performance was compared with alternative depletion methods rather than assessed in isolation.
- Nanopore sequencing offers the possibility of analysing data while sequencing is still in progress.
Limitations
The method still depends on microbial growth in blood culture. It is therefore not a truly culture-independent diagnostic test.
Organisms that grow slowly, inconsistently or only at low abundance may remain difficult to detect. Previous antimicrobial exposure could also reduce culture enrichment.
The analytical threshold was selected within the study dataset and requires independent validation. The authors additionally observed reduced sequence yield with some high-GC organisms.
Genotypic resistance prediction is not equivalent to phenotypic susceptibility testing. This is particularly important in fungi, where recognised resistance markers explain only part of the observed susceptibility phenotype.
Why might this matter for fungal diagnostics?
The most immediate fungal application would be rapid analysis of blood cultures containing Candida or other yeasts. Faster species identification and detection of recognised resistance determinants could reduce the interval between blood-culture positivity and targeted antifungal treatment.
The relevance to invasive mould disease is more limited because blood cultures are usually negative in invasive aspergillosis and most other mould infections.
Diagnostic interpretation
This is a credible translational sequencing workflow with potential relevance to yeast bloodstream infection. It is not yet a replacement for culture, susceptibility testing or specialist interpretation.
Current practice impact: No immediate change.
Future potential: High for rapid identification and resistance analysis of yeasts recovered from blood cultures.
2. Removing human DNA to improve plasma metagenomic sequencing
Zhan S, Zheng Y, Wu T, et al. Nucleosome-targeted host DNA depletion enables automated plasma metagenomic sequencing for sensitive detection of bloodstream pathogens. Journal of Translational Medicine. 2026.
Diagnostic development: Automated host-DNA depletion before plasma metagenomic sequencing.
Technology-watch rating: ★★★ High priority
What did the researchers study?
Plasma metagenomic sequencing is attractive because it can detect pathogen DNA without requiring microbial growth. However, microbial DNA is usually overwhelmed by human cell-free DNA, reducing sensitivity and increasing sequencing requirements.
The researchers developed an automated host DNA-depletion mNGS assay, integrating:
- nucleosome-targeted removal of human DNA;
- automated DNA extraction;
- automated library preparation;
- and metagenomic sequencing.
Analytical validation included limit of detection, linearity, precision and contamination control. Clinical performance was then assessed in 107 patients with suspected bloodstream infection.
The method was compared with:
- blood culture;
- conventional microbiological testing;
- standard mNGS without host depletion;
- and a composite clinical reference standard.
What did they find?
Nucleosome-targeted depletion reduced the human-DNA background by an average of 66-fold and increased the proportion of microbial sequencing reads by approximately 46.7-fold.
The reported limits of detection were:
- 9.1–38 genome equivalents/mL for bacteria and fungi;
- 283–321 genome equivalents/mL for viruses.
The assay showed good linearity across tested concentrations, with reported R² values between 0.915 and 0.989.
Automation maintained strong quantitative correlation with the corresponding manual protocols. It also reduced common skin contaminants by 71.7% and environmental contaminants by 83.7%.
In the 107-patient clinical cohort, the host-depleted assay achieved:
- a pathogen-detection rate of 64.49%;
- clinical positive percentage agreement of 95.24%;
- and total percentage agreement of 88.79%.
The authors reported significantly better pathogen detection than standard mNGS, blood culture and conventional microbiological testing.
The method was particularly effective for organisms associated with very low concentrations of circulating DNA, including Mycobacterium tuberculosis and Rickettsia species.
Strengths
- The study addresses one of the central technical barriers to plasma mNGS.
- It included both analytical and clinical validation.
- The clinical cohort was substantially larger than many early diagnostic-development studies.
- Automation could improve reproducibility and reduce variation between operators.
- Contamination reduction is particularly important for highly sensitive sequencing methods.
- Fungi were included in the analytical limit-of-detection experiments.
Limitations
A composite clinical reference standard can introduce classification uncertainty, particularly where no single definitive comparator exists.
The abstract does not provide organism-by-organism performance. The overall positive percentage agreement may therefore conceal variation between bacteria, fungi and other pathogen groups.
Improved analytical sensitivity does not automatically establish clinical specificity. Detection of microbial DNA may reflect active infection, transient DNA release, contamination or organisms present at another anatomical site.
The study needs replication in independent populations and evaluation in patient groups at high risk of invasive fungal disease.
Why might this matter for fungal diagnostics?
This study is directly relevant to the goal of non-invasive fungal detection.
Blood cultures have poor sensitivity for invasive aspergillosis and many other invasive mould infections. Plasma mNGS could potentially identify circulating fungal DNA without bronchoscopy or tissue biopsy.
The reported fungal limits of detection are encouraging, but key clinical questions remain:
- Which fungal species were included in analytical validation?
- How well did the test detect proven or probable invasive fungal disease?
- Was sensitivity maintained after antifungal treatment had started?
- Could the assay distinguish invasive infection from contamination or colonisation?
- How did performance compare with galactomannan, β-D-glucan and targeted fungal PCR?
- Did sequencing results alter treatment or improve outcomes?
Diagnostic interpretation
This is the most directly relevant Technology Watch study this week. It demonstrates that selective depletion of nucleosome-bound human DNA can substantially improve microbial signal recovery from plasma while reducing contamination and manual processing.
The findings do not yet establish plasma mNGS as a routine test for invasive fungal disease, but they remove an important technical obstacle.
Current practice impact: No immediate change.
Future potential: High for culture-independent detection of invasive fungal infection, particularly if fungal-specific clinical validation confirms the analytical findings.
3. Artificial intelligence only shortens turnaround when the workflow allows results to be released
Davidson R, Heinstein C, Patriquin G, et al. Improving turnaround times with artificial intelligence in microbiology. Journal of Clinical Microbiology. 2026.
Diagnostic development: AI-assisted interpretation and release of culture results.
Technology-watch rating: ★★ Worth watching
What did the researchers study?
This dual-centre Canadian study evaluated the introduction of PhenoMATRIX, an AI-based system used with microbiology laboratory automation.
The software provides continuous sorting and interpretation support for urine cultures. The study compared turnaround times before and after implementation in:
- a lower-volume tertiary hospital laboratory;
- and a high-volume community laboratory.
What did they find?
At both sites, the AI system made interpretable results available earlier. However, earlier availability did not automatically produce earlier reporting.
At the tertiary hospital, implementation of the AI system alone was initially associated with a longer time to result reporting because results became available but were not released promptly.
When PhenoMATRIX Plus was introduced to permit automated release of negative results, reporting time improved by approximately 1.3 hours.
At the community laboratory, automated result release was not used. Instead, manual screening was moved from 16:00 to 08:00 so staff could review AI-generated results earlier. This reduced reporting time by approximately 5.3 hours.
The system also reduced the hands-on time required for laboratory staff to interpret and finalise urine-culture results.
Strengths
- The study involved two laboratories with different workloads and operating models.
- It assessed implementation under real laboratory conditions.
- It distinguished between result availability and actual result reporting.
- It demonstrates that AI performance cannot be evaluated separately from laboratory workflow.
- Benefits were achievable through either automated release or earlier human review.
Limitations
The study involved bacterial urine cultures rather than fungal diagnostics.
The improvements depended partly on local operational changes. The larger reduction at the community laboratory followed earlier manual review, not autonomous AI reporting alone.
The study therefore evaluates an AI-enabled laboratory system rather than the independent diagnostic accuracy of an algorithm.
Automated release is also easiest for clearly defined negative or uncomplicated results. More complex cultures will continue to require human review.
Why might this matter for fungal diagnostics?
AI could eventually support fungal laboratories by:
- screening culture plates for growth;
- identifying likely negative plates;
- detecting early colony development;
- highlighting mixed or unusual morphologies;
- prioritising plates for expert review;
- and reducing repetitive manual examination.
However, fungal cultures are generally more complex than routine urine cultures. Growth can be slow and variable, and clinically important colonies may initially be subtle.
Fungal culture interpretation may also depend on specimen type, colony quantity, patient context and differentiation between colonisation and disease.
Diagnostic interpretation
The most important lesson is operational:
AI does not shorten diagnostic turnaround simply by generating an earlier interpretation. Laboratories must also redesign result-review and reporting processes.
For fungal diagnostics, AI is likely to enter initially as a triage and workflow tool rather than an autonomous reporting system.
Current practice impact: None for fungal diagnostics.
Future potential: Moderate for automated culture screening, prioritisation and release of clearly defined results.
4. AI-supported microscopy improves performance when paired with expert confirmation
Feoktistov I, Koelewijn R, Tieken C, et al. Development and validation of the AI-predictive ParaScout in-vitro diagnostic system for the microscopic detection of gastrointestinal helminths in stool. Emerging Microbes & Infections. 2026;15(1).
Diagnostic development: Digital microscopy combined with deep-learning-assisted image recognition.
Technology-watch rating: ★★ Worth watching
What did the researchers develop?
The investigators developed ParaScout, an in-vitro diagnostic system combining:
- an affordable commercially available digital microscope scanner;
- automated image capture;
- a cloud-based deep-learning algorithm;
- and expert confirmation of structures highlighted by the system.
The algorithm had been trained to recognise 15 gastrointestinal helminth species.
How was it evaluated?
Performance was assessed using 50 validated stool specimens examined by four expert technicians.
Forty-five samples contained one or more helminths. Across those samples, 63 individual species identifications were possible. Because four technicians examined each sample, manual examination could theoretically have produced 252 correct identifications.
What did they find?
Manual microscopic examination produced:
- 16 false-negative results;
- 11 false-positive results;
- sensitivity of 93.7%;
- and specificity of 99.6%.
When ParaScout was used to highlight suspected structures for human expert confirmation, performance increased to:
- sensitivity of 98.8%;
- specificity of 99.9%.
The best performance therefore came from an AI-plus-expert model, rather than fully autonomous interpretation.
Strengths
- The system was evaluated against multiple expert technicians.
- It included both negative and positive specimens.
- The study assessed a practical diagnostic workflow rather than image classification alone.
- The use of expert confirmation substantially limited false-positive reporting.
- The equipment was designed around an affordable commercial microscope scanner.
- The study included multiple target species rather than a single organism.
Limitations
The validation set was relatively small, involving only 50 specimens.
The system was tested on organisms included in its training set. Performance with rare species, unusual morphology, damaged structures or unexpected artefacts remains uncertain.
Helminth eggs and larvae are often larger and more morphologically distinctive than fungal hyphae, yeast cells or spores. Direct transfer of the reported performance to fungal microscopy cannot be assumed.
Cloud-based analysis may also raise practical questions concerning connectivity, data governance, cybersecurity and service continuity.
Why might this matter for fungal diagnostics?
Microscopy remains important in the diagnosis of fungal infection, but performance is influenced by operator experience, specimen preparation, staining quality and organism burden.
A similar AI-assisted approach might eventually support recognition of:
- fungal hyphae;
- budding yeast cells;
- pseudohyphae;
- fungal elements in tissue;
- calcofluor-white-positive structures;
- characteristic morphology in direct specimens;
- and possible contaminants or artefacts.
The most plausible model would be similar to ParaScout: the algorithm highlights suspicious structures and a trained scientist confirms or rejects them.
Diagnostic interpretation
This study provides a useful demonstration of how AI microscopy may be introduced safely. The algorithm did not eliminate the expert. Instead, it improved sensitivity by directing expert attention to potentially important structures.
That model is likely to be more realistic for fungal microscopy than immediate autonomous reporting.
Current practice impact: None for fungal diagnostics.
Future potential: Moderate for screening and decision support, especially in laboratories with limited specialist microscopy capacity.
Overall assessment
The most important paper this week is the study of nucleosome-targeted host-DNA depletion before plasma mNGS.
A 66-fold reduction in human-DNA background, approximately 47-fold enrichment of microbial reads, low analytical detection limits for bacteria and fungi, and strong clinical positive percentage agreement suggest that host-DNA depletion may substantially improve plasma sequencing.
The crucial next step is fungal-specific clinical validation.
The same-day culture-enriched sequencing paper is more immediately applicable to bloodstream yeasts. It could shorten species identification and resistance analysis after a blood culture becomes positive, but it remains dependent on microbial growth.
The two AI studies reinforce a consistent message: AI produces the greatest diagnostic benefit when embedded within a carefully designed human workflow.
In the urine-culture study, earlier AI interpretation only improved reporting after the laboratory changed its result-release processes. In the microscopy study, performance was best when AI highlighted structures for expert confirmation.
What should fungal diagnostic laboratories take from this?
No immediate change to routine testing is justified.
The developments worth monitoring are:
- plasma host-DNA depletion for fungal cell-free DNA detection;
- fungal-specific clinical validation of plasma mNGS;
- rapid sequencing of yeast-positive blood cultures;
- sequence-based antifungal-resistance prediction;
- AI-supported fungal culture screening;
- and AI-assisted microscopy with expert confirmation.
The key question is not simply whether these technologies can detect microbial material. It is whether they can produce reproducible, clinically interpretable results that improve diagnosis, treatment decisions or patient outcomes beyond existing culture, histopathology, antigen testing, serology and targeted molecular assays.
References
- Davidson R, Heinstein C, Patriquin G, et al. Improving turnaround times with artificial intelligence in microbiology. Journal of Clinical Microbiology. 2026. PubMed
- Feoktistov I, Koelewijn R, Tieken C, et al. Development and validation of the AI-predictive ParaScout in-vitro diagnostic system for the microscopic detection of gastrointestinal helminths in stool. Emerging Microbes & Infections. 2026;15(1). PubMed
- Sajib MSI, Oravcová K, Brunker K, et al. Rapid and modular workflows for same-day sequencing-based detection of bloodstream infections and antimicrobial resistance determinants using culture-enriched samples. Microbiology Spectrum. 2026;14(7):e0324025. PubMed
- Zhan S, Zheng Y, Wu T, et al. Nucleosome-targeted host DNA depletion enables automated plasma metagenomic sequencing for sensitive detection of bloodstream pathogens. Journal of Translational Medicine. 2026. PubMed