Fungal metatranscriptomics moves from descriptive ecology toward predictive science

Fungi are everywhere. They thread through soils, litter, freshwater, ocean sediments, plant roots, and animal guts, and their activities help sustain nearly every ecosystem on Earth. Yet only about 200,000 fungal species have been formally described, while estimates of true diversity range from 2.2 to 3.8 million. Understanding what these organisms are actually doing in nature, not merely which ones are present, has long been a central challenge for ecologists. A new review published in IMA Fungus argues that metatranscriptomics, the direct sequencing of RNA transcripts recovered from environmental samples, is now the most powerful route to answering that question, provided the field can overcome significant computational and database hurdles.

Traditional fungal community studies have relied heavily on amplicon sequencing of the internal transcribed spacer (ITS) region or other rRNA genes. This approach has generated decades of valuable data, but it carries well-documented constraints. Universal primers such as ITS1F and ITS2 amplify different fungal groups with varying efficiency, potentially skewing estimates of community composition between phyla like Ascomycota and Basidiomycota. rRNA gene copy numbers also vary enormously among species, from roughly 100 to 200 copies in the yeast Saccharomyces cerevisiae to more than 500 copies in some Basidiomycetes, complicating any attempt to infer abundance from amplicon data. Most importantly, amplicon sequencing reveals nothing about function. It identifies which fungi are present but not which genes they are actively expressing.

Metatranscriptomics sidesteps these problems by capturing actively transcribed genes in situ. Because it sequences RNA directly, it circumvents primer bias and reveals the transcripts present under natural conditions. The review highlights striking examples of what this makes possible. In the rhizosphere of Norway spruce, researchers found that Cortinarius fungi actively expressed transcripts related to the degradation of aromatic and heterocyclic organic macromolecules, a core metabolic function that amplicon sequencing could not have identified. In Chinese coastal waters, integrating ITS amplicon data with metatranscriptomics revealed that planktonic Dikarya fungi contributed more than 3% of eukaryotic carbohydrate-active enzyme (CAZyme) transcripts, disclosing their in situ role in mediating marine organic carbon degradation. Metatranscriptomics thus allows researchers to simultaneously determine who is present and what they are doing, linking taxonomic identity with functional activity in a single dataset.

A standard eukaryotic metatranscriptomic workflow begins with sample collection and total RNA extraction from environmental sources such as plant tissues, soil, litter, or water. Because fungal RNA typically represents only a minor fraction of total environmental RNA, often less than 5%, the extracted RNA undergoes ribosomal RNA depletion and poly-A enrichment to concentrate the eukaryotic messenger RNA fraction. This enriched mRNA is reverse-transcribed into complementary DNA for library construction, with sequencing typically performed on Illumina platforms. Raw reads then undergo quality control, adapter trimming, and removal of low-quality sequences and potential host RNA before custom bioinformatic pipelines are applied to separate host and fungal sequences, classify fungal taxa, and characterize functions.

The review systematically categorizes current analytical workflows into two main computational strategies: assembly-based and alignment-based methods. Assembly-based pipelines first extract target reads by mapping the qualified dataset onto comprehensive reference databases such as RDP and SILVA, then apply de novo assemblers like Trinity, Oases, or rnaSPAdes to reconstruct longer contigs. These assembled sequences provide more phylogenetically informative sites, enabling differentiation of closely related species and even discovery of novel taxa absent from existing databases. In spruce rhizosphere samples, an assembly-based pipeline successfully identified over 95.5% of fungal transcripts at the phylum level and 34% at the species level. Trinity, the most widely used transcriptome assembler, employs a three-module architecture (Inchworm, Chrysalis, and Butterfly) and achieves notably high completeness in fungal gene recovery. In a mouse gut metatranscriptomic dataset, Trinity generated 21,454 annotatable contigs and recruited 50.3% of raw reads, whereas the generic metagenomic assembler MetaVelvet yielded only 13,706 contigs and mapped merely 32.8% of reads.

Alignment-based pipelines, by contrast, bypass the computationally intensive assembly step by directly matching sequencing reads to existing reference databases. These tools fall into three categories: genome-based classifiers such as Kraken2 and Centrifuge, which use optimized k-mer or FM-index approaches for ultra-fast assignment; marker gene-based methods such as MetaPhlAn4, EukDetect, and MicroFisher, which exploit species-specific marker genes; and protein family-based methods such as FunOMIC-P. Standardized comparisons show that alignment-based pipelines are fast and scalable, with fungal-specific marker-based tools performing better than general-purpose classifiers. In synthetic mycobiome evaluations, FunOMIC identified the majority of fungal species while EukDetect produced community profiles most closely resembling the true composition. Notably, the accuracy of these marker-based tools remained largely unaffected even when bacterial reads constituted 90 to 99% of the sample, suggesting robustness to bacterial background noise that k-mer classifiers may lack. However, alignment-based performance depends heavily on database comprehensiveness, and short reads often preclude precise species-level classification, particularly for highly conserved genomic regions.

For functional characterization, the two strategies again diverge. Assembly-based approaches reconstruct transcripts, predict open reading frames with tools like TransDecoder, and annotate predicted proteins against specialized databases including KOG, GO, KEGG, CAZy, and Pfam. Integrated pipelines such as Trinotate combine BLAST, HMMER, and InterProScan to achieve protein domain identification, GO term assignment, and KEGG pathway mapping simultaneously. This strategy enables quantitative analysis of assembled transcripts across conditions and allows tracing of functional activities back to specific fungal taxa. In forest soils, for example, an assembly-based approach demonstrated that saprotrophic fungi significantly upregulate plant cell wall-degrading enzyme genes while ectomycorrhizal fungi highly express nitrogen transporter genes, clearly elucidating the ecological roles of major fungal guilds at refined taxonomic levels. Alignment-based functional classification, using tools like FMAP and HUMAnN2, maps raw reads directly to reference databases such as KEGG, SEED, and UniRef, offering exceptional efficiency but suffering from short-read constraints and an inability to distinguish whether conserved metabolic genes originate from fungi or bacteria.

Despite these advances, major hurdles persist. Fungal transcriptome complexity presents multidimensional challenges: rust fungi have average genome sizes of 300 megabases ranging up to 900, characterized by high content of repetitive genes, transposable elements, and introns. Magnaporthe oryzae generates 4,270 alternatively spliced isoforms from 2,413 genes, with roughly one-third detected only during host infection. Database biases compound these difficulties, with over 80% of sequenced fungal genomes falling within Ascomycota and Basidiomycota while early-diverging lineages remain severely underrepresented. General-purpose databases exhibit significant underrepresentation of unique fungal genomic features such as secondary metabolite gene clusters and carbohydrate-active enzymes involved in lignocellulose degradation, with over 60% of polyketide synthase genes lacking accurate annotation in KEGG. The absence of standardized analytical workflows further compromises comparability and reproducibility across studies.

Looking forward, the review identifies several promising directions. Long-read sequencing platforms such as PacBio enable direct acquisition of full-length transcripts, permitting accurate resolution of complex fungal gene models. In the Zymoseptoria tritici genome, Iso-Seq technology precisely delineated 5′ and 3′ UTR boundaries for 73% of re-annotated gene models and uncovered prevalent alternative splicing mechanisms previously unresolved. Expanding fungal-specific genomic and functional databases, driven by initiatives like the JGI 1000 Fungal Genomes Project with over 3,800 high-quality genomes currently released, will fundamentally enhance alignment-based strategies. Specialized resources such as FungiDB, integrating data from more than 1,000 fungal species with 40% higher GO annotation coverage than general-purpose databases, and MycoCosm, focused on environmentally relevant fungi, are progressively addressing database gaps. Artificial intelligence is emerging as a pivotal tool, with deep learning-based gene prediction tools improving ORF prediction accuracy by 18% and AlphaFold3 capable of accurately predicting fungal enzyme structures of unknown genes. Containerization technologies such as Nextflow and Snakemake are simplifying deployment and reproducibility of complex analytical processes, while community-driven benchmarking initiatives like CAMI promote continuous tool improvement through standardized test datasets.

Collectively, these developments signal a transition in fungal metatranscriptomics from descriptive ecology toward a predictive science. By simultaneously resolving taxonomic identity and functional activity, metatranscriptomics overcomes the limitations of traditional amplicon sequencing and enables in situ analysis of fungal community functions across diverse ecosystems. As long-read sequencing matures, fungal-specific databases expand, and artificial intelligence refines annotation and classification, researchers will gain increasingly powerful support for understanding and managing fungal resources. The capacity to predict, rather than merely describe, how active fungal communities shape carbon cycling, nutrient fluxes, and plant performance could transform how scientists approach ecosystem management in an era of rapid environmental change.

Subject of Research: Fungal metatranscriptomics

Article Title: Interpreting fungal ecological contributions through taxonomic and functional profiling of metatranscriptomics

Article References: Hu, W., Hu, J., Wang, Z., & Wang, H. (2026). Interpreting fungal ecological contributions through taxonomic and functional profiling of metatranscriptomics. IMA Fungus, 17, Article e185187. https://doi.org/10.3897/imafungus.17.185187

Image Credits: AI Generated

DOI: 10.3897/imafungus.17.185187

Keywords: metatranscriptomics, mycology, bioinformatics, fungal ecology, genomics, Interpreting, fungal, ecological, contributions, taxonomic, functional, profiling