Resistome predictors are laborious initiatives only justified when provide evident clinical benefits. AMR bacteria show high virulence and national cost burden, associated with 5 million deaths and 1.3 direct causality in 2019 (Antimicrobial Resistance Collaborators, 2022). In 2017, the World Health Organization (WHO) presented a list with the top-12 bacterial pathogens (Tacconelli et al., 2018), which are extensively covered by resistome tools.

Fungal infections amount to 1.6 million deaths in developing countries (Almeida et al., 2019). Cryptococcus neoformans, Candida auris, Candida albicans, and Aspergillus fumigatus conform the critical priority group in the recently released Fungal Priority Pathogen List by the WHO.¹ An increment of candidiasis caused by different Candida species has been observed (Lamoth et al., 2018), particularly C. auris in Europe (Kohlenberg et al., 2022). Fungal pathogens are gaining relevance in risk groups due to aging, AIDS, cancer chemotherapy, organ-transplanted patients, viral infections (flu and COVID-19), or ICU admission. Some cases can be deadly, such as systemic fungemia, pneumonia, and meningitis if not treated timely and properly. Candida is the fourth pathogen more often found in sepsis in the United States (Delaloye and Calandra, 2014).

Mortality and morbidity associated to AMR fungi is increasing (Perlin et al., 2017; Fisher et al., 2022; Gow et al., 2022; Figure 1A). The antimicrobial control of these infections is compromised by the small drug arsenal and genomic plasticity. Fungi are eukaryotic, i.e., molecularly similar to humans, which limits the number of antifungal family drugs to five. Only azoles, echinocandins, and polyenes are used to treat systemic infections. Genome plasticity facilitates the adaptive resistance to chemotherapeutical options during long treatments. Multidrug resistant phenotypes are also detected (Pfaller et al., 2012; Pham et al., 2014). In contrast to adaptive resistance, intrinsic resistance, caused by naturally occurring ancestral mutations, has also been detected (Chowdhary et al., 2018). Especially worrisome are the intrinsic pan-resistance observed in Lomentospora prolificans (Wu et al., 2020) and the emergence of fluconazole resistant Candida parapsilosis besides its intrinsic higher MICs to echinocandins (Trevijano-Contador et al., 2022). The therapeutic failure using the two preferred drug classes forces the prescription of alternative treatments such as amphotericin B, the antifungal with a broadest spectrum.

Fungal infections are usually treated by empirical therapy based on historical records or after applying phenotypical methods. Antifungal susceptibility testing involves the standardized exposition of the fungus to different concentrations of drug in liquid or solid media using reference protocols (CLSI and EUCAST). Therapeutic efficacy can be therefore predicted by quantitatively assessing the minimum inhibitory or effective concentrations. Clinical break points for some antifungals and species are available. Disadvantages of phenotypical methods include low reproducibility by subtle operational changes and potential absence of cutoff points. Besides, they are labor-intensive and time-consuming which has a direct impact in mortality (Garey et al., 2006; Chamilos et al., 2008).

Alternatively, the identification of mutations by sequencing outperforms MIC for antifungal therapy success (Shields et al., 2012; Alexander et al., 2013). However, only a limited number of genes can be sequenced by conventional Sanger sequencing or screened by real-time PCR (Durand et al., 2021). Resistance marker expression can be approached by MS MALDI-TOF (Vella et al., 2017; Maenchantrarath et al., 2022), which shows short turn-around-times but lacks of standards for universal application (Durand et al., 2021).

The polygenic nature of antifungal resistance (Gonçalves et al., 2016) favors the utilization of massive techniques. Cost-effective new generation sequencing (NGS) alone or complemented with better genome assemblies achieved by third-generation sequencing (TGS) facilitates obtaining whole-genome sequences. Genomics provides a view of the strain genetics at the highest, nucleotide-level, resolution, and the simultaneous evaluation of a multiplicity of factors (Garnaud et al., 2015; McTaggart et al., 2020; Schikora-Tamarit and Gabaldón, 2022b).

A central question is whether the current scientific knowledge includes enough fungal AMR mechanisms to fuel the development of efficient resistome predictors. Laboratory and clinical studies have indeed described numerous adaptive and intrinsic antifungal AMR processes with molecular precision.

Many point mutations linked to AMR have been characterized, which would be akin to the “variant mode” of bacterial resistome predictors. These include those in primary antifungal targets: the key enzyme for ergosterol biosynthesis sterol 14α demethylase for azoles, and glucan synthase for echinocandins. Many of these mutations produce residue changes in relevant (hot-spot) zones like drug binding pockets (Garcia-Effron et al., 2009). Mutations in non-primary target proteins can also produce resistance, such as in the ergosterol biosynthesis pathway to azoles (Alcazar-Fuoli and Mellado, 2012). Loss-of-function (LOF) resistance mutations, such as premature stop codons and insertion of transposable elements, have been described for instance in the ergosterol metabolism enzyme ERG3 (Lupetti et al., 2002; Ksiezopolska et al., 2021). Intrinsic AMR examples are the numerous polymorphisms in ERG11 and FKS proteins, causing azole and echinocandin resistance in C. auris (Muñoz et al., 2018) and C. parapsilosis (Garcia-Effron et al., 2008), respectively. Decrease of the intracellular drug concentration to ineffective levels can be achieved by gain-of-function (GOF) mutations in positive regulators of antifungal efflux pumps (Cannon et al., 2009). AMR phenotypes are favored in isolates with hypermutator behavior due to altered DNA repair systems (Healey et al., 2016). Resistant phenotypes caused by point mutations are usually only observed in diploid fungi when the sensitive allele is lost (Garnaud et al., 2015).

Expression-driven resistance can also be obtained by: (i) mutations in response elements of cis promoters of efflux pumps (Gaur et al., 2004); (ii) new promoters introduced by transposable elements (Hu et al., 2021); (iii) longer mRNA stability by enzymatic variants that increase polyadenylation (Manoharlal et al., 2010); and (iv) LOF mutations of lysine-acetylation enzymes of histones via epigenetic chromatin modification (Orta-Zavalza et al., 2013; Tscherner et al., 2015).

Fungal AMR often involve large chromosomal changes affecting copy number and expression regime of genes. Large changes include ploidy alterations, the combination of chromosomes into new ones, and the generation of isochromosomes (Selmecki et al., 2006, 2008; McTaggart et al., 2020). For example, chromosomal monosomies have been associated with tolerance to fluconazole and 5-flourocytosine (Yang et al., 2013; Uddin et al., 2021). The copy number of genes coding primary targets can be also increased by local duplication (Chow et al., 2020).

An essential issue for routine antifungal resistome is how many fungal AMR markers are detectable by WGS and with which degree of automation.

Intrinsic resistance involves precise taxon identification. For that, k-mers thresholds for species assignation have been reported (Gostinčar, 2020). Precise sub-species detection requires the application of phylogenomic tools or the standard sequence type, ST, scheme (Taylor and Fisher, 2003).

Point polymorphisms associated to AMR follow the classical “variant-mode” resistome principles. Illumina NGS technology meets quality and coverage standards enough to identify LOF and GOF residue changes, premature stop codons, small indels, and gene duplications (Garnaud et al., 2015). Protocols such as OVarFlow and SnpEff are available to call and study point and small indels in eukaryotic genomes (Bathke and Lühken, 2021; Cingolani, 2022). Fungal-specific and other tools have been already designed for this task, such as YMAP (Abbey et al., 2014). Deep coverages are also adequate to find heteroresistant subpopulations (Zhai et al., 2022). Homozygosity (~100% reads) and heterozygosity (~50% reads) are approachable using the proportion of mutated reads (Spettel et al., 2019).

Large chromosomal changes are harder to parametrize although they can be evaluated through tools like PerSVade (Schikora-Tamarit and Gabaldón, 2022a) or Assemblytics (Nattestad and Schatz, 2016). NGS permits proportional read count alteration in certain chromosomes or in the vicinity context of marker genes. Transposable elements are identifiable using sequence profiles from databases like Repbase (Kapitonov and Jurka, 2008). However, genome environment modifications through transposons and chromosomal rearrangements may require high-quality assemblies as those provided by TGS platforms.

The nature of some antifungal resistance modes is still too opaque to be investigated. The resulting alteration profiles by genomics can be a mixture of AMR causal, linked, suspected, or irrelevant changes. Players include drug targets, regulators, pathways, promoters, mRNA polyadenylation, and lysine acetylases. To assess the relevance of sequence variations, the number of sequenced genomes and the molecular knowledge of fungal pathogens should be strengthened (Figure 1B).

Beyond performance in the research environment, the ultimate goal of antifungal resistome predictors is their clinical application. Full advantage of genome-based tools has been taken for bacterial pathogens in the clinical laboratory in the recent years (Deurenberg et al., 2017).

Fungal resistome tools should be economically efficient, in particular in developing countries. The number of samples per NGS run should be optimized. The number of reads and assemblies obtained by NGS should satisfy quality metrics such as high genome coverage and a small contig number. Concerning TGS, nanopore flongle cells or multiplexing show lower fidelity but are cost-effective options for NGS-TGS hybrid approaches (Lipworth et al., 2020) or utilization in low- and middle-income countries. Expense can also be reduced by targeted resequencing (Spettel et al., 2019).

A principal hurdle in genomics is the downtime to obtain sequencing results. Long turnaround-times may be assumable for chronic non-critical superficial infections (like candiduria), but not for life-threatening ones (like fungemia, pneumonia or meningitis) that could be mortal within 24 h. The later demands in-house facilities as dependence of third externalized parties can substantially delay outcomes (Raven et al., 2018). Thus, the initial investment to purchase NGS facilities would be eventually profitable. Furthermore, nanopore technologies are relatively cheap and can be escalated up to 48 simultaneous cells.

The requirement of a high bioinformatic knowledge may also hamper the predictor utilization. Unexperienced laboratory staff should be trained to execute and interpret the provided software. On-line resources that provide human readable reports, such as Pathogenwatch² and Microreact (Argimón et al., 2016), besides handy computational environments like Galaxy (Galaxy Community, 2022) may facilitate analyses for laboratories worldwide.