Species-Level Taxonomy vs Functional Metagenomics: Choosing the Right Microbiome Strategy

When designing a microbiome study, research teams face a fundamental question: how much biological information does the project actually need? Some studies require broad whole-metagenome sequencing to investigate microbial genes, pathways, mobile genetic elements, or previously uncharacterized genomes. Others primarily need to determine which microbial species are present, how their relative abundances differ among research groups, or which taxa are associated with an experimental phenotype.

More sequencing does not automatically produce a better-designed study. Deep shotgun metagenomics can provide valuable genome-wide information, but it also increases sequencing and bioinformatic requirements, and its effective microbial depth can decline sharply in samples dominated by host DNA. Conversely, a taxonomy-focused strategy may answer a species-level research question more efficiently while preserving resources for additional samples, technical controls, or independent validation. This guide uses a deliverable-first framework to help research teams decide when species-level taxonomy is sufficient, when functional inference may support exploratory analysis, and when direct shotgun metagenomics is necessary.

Direct Answer: What Output Do You Actually Need?

The most appropriate sequencing strategy should be selected from the required primary deliverable rather than from the perceived sophistication of a particular technology.

• Species-Level Taxonomic Profiling — "Who is there?": If the primary objective is to compare species composition, identify taxa associated with experimental conditions, characterize community diversity, or develop candidate microbial signatures for research, species-level profiling may be sufficient. Depending on sample quality and study design, potential approaches include reduced-representation metagenomics such as 2bRAD-M, full-length marker-gene sequencing, or shallow shotgun metagenomics.
• Predicted Functional Potential — "What functions might be represented?": Marker-gene data can be used with tools such as PICRUSt2 to predict functional potential from reference genomes. Such predictions are useful for exploratory hypothesis generation but should not be treated as equivalent to direct measurement of microbial genes.
• Direct Gene and Pathway Profiling — "What genes are actually represented in the sequenced DNA?": When the project requires direct evidence for gene families, pathway components, antibiotic resistance genes, or other genomic features that may vary within a species, shotgun metagenomics is generally more appropriate than taxonomic inference alone.
• Genome Reconstruction — "Can microbial genomes be reconstructed from the community?": Recovery of Metagenome-Assembled Genomes (MAGs), investigation of novel genomic content, and genome-resolved analysis require substantially greater effective microbial coverage and should be treated as distinct objectives from community-level taxonomic profiling.

Key Takeaways

• Start with the Deliverable: Define whether the required output is a species abundance table, predicted functional profile, directly observed microbial gene profile, or reconstructed microbial genomes.
• Species Taxonomy Can Be Sufficient for Many Association Studies: When the research hypothesis concerns differences in community composition or the abundance of particular taxa, additional genome-wide sequencing may not be necessary to answer the primary question.
• Predicted Function Is Not Direct Functional Measurement: Reference-based tools can estimate potential functions, but predictions depend on reference genomes and cannot reliably capture every strain-specific gene, horizontal gene transfer event, plasmid, or previously uncharacterized sequence.
• Sequencing Depth Is Endpoint-Dependent: The depth needed for reference-based taxonomy can be substantially lower than the depth required for broad protein coverage or de novo MAG reconstruction.
• Host DNA Changes the Calculation: In host-rich specimens, untargeted shotgun reads may be dominated by host sequences, reducing the effective microbial information obtained at a given total sequencing depth.
• DNA Measures Genomic Potential, Not Activity: Metagenomic DNA can indicate which genes are present. Metatranscriptomics provides information about microbial transcription, while metabolomics can add information about measured metabolites and biochemical state.

The Four Microbiome Deliverables: Defining the True Analytical Endpoint

A useful way to plan a project is to separate microbiome outputs into four increasingly information-intensive levels. Each level answers a different biological question and requires different amounts and types of sequence information.

The distinction between these deliverables is important because a species-level abundance table and a microbial gene catalog answer different questions. A project designed to ask whether the abundance of Akkermansia muciniphila differs between experimental groups does not automatically require reconstruction of its entire genomic content. Conversely, detecting a species does not establish whether a particular strain-specific gene, plasmid, or resistance determinant is present.

For projects requiring direct taxonomic and functional metagenomic analysis, explore our microbial metagenomics analysis platform.

When Species-Level Taxonomy May Be Enough

Species-level taxonomy can be an appropriate primary endpoint when the biological hypothesis is fundamentally compositional. Examples include determining whether particular microbial species are enriched or depleted across experimental groups, assessing whether community structure changes after an intervention, identifying reproducible taxa associated with an environmental condition, or building research classifiers from microbial abundance features.

In these situations, the essential statistical input may be a well-controlled species abundance matrix rather than a comprehensive catalog of every microbial gene. Generating substantially more sequence data than the analysis requires does not automatically strengthen the biological conclusion. Once sufficient taxonomic information and technical reproducibility have been achieved, study power may instead benefit from adequate biological replication, balanced cohort design, appropriate controls, and independent validation.

Species-level profiling can be particularly attractive for difficult samples where full shotgun sequencing is inefficient. Host-rich tissues, low-biomass specimens, and degraded archival material can contain relatively little recoverable microbial DNA compared with host DNA. In these settings, a taxonomy-focused assay may provide a more practical route if genome-wide functional information is not a primary requirement.

When Species-Level Taxonomy Is Not Enough

Species identity should not be treated as a substitute for genomic information when the research question depends on variation that occurs within species or moves between genomes. Closely related strains can differ substantially in gene content even when they receive the same species-level taxonomic label.

Direct metagenomic sequencing becomes more important when the project asks questions such as:

• Which antibiotic resistance genes are present? Resistance determinants can be gained or lost independently of species identity and may occur on mobile elements.
• Which plasmids, prophages, or mobile genetic elements are present? These features cannot be inferred reliably from a species abundance table.
• Does a microbial species contain a particular metabolic gene cluster? Strain-level gene content may differ substantially from the reference genome used for taxonomic assignment.
• Are previously uncharacterized genes or organisms present? Reference-dependent taxonomic methods are intrinsically limited by the available database.
• Is genome reconstruction required? MAG assembly depends on sufficient sequence coverage distributed across microbial genomes, not simply on confident species identification.

For these endpoints, direct sequence evidence is more informative than taxonomic prediction. The correct question is therefore not whether taxonomy or metagenomics is universally "better," but whether the study needs organism identity, direct genomic content, or both.

Functional Redundancy: Why Taxonomic and Functional Profiles Are Different

Microbial community composition and functional potential are related but not interchangeable. Different organisms can encode overlapping metabolic capabilities, while closely related strains can differ in accessory genes and pathway components. As a result, communities with different taxonomic compositions may sometimes retain similar broad functional capabilities.

Louca et al. demonstrated this decoupling between taxonomic and functional structure in a global ocean microbiome study (Louca et al., 2016). Their findings provide an important ecological example of why taxonomic turnover does not always produce equivalent changes in broad functional composition. The magnitude and biological meaning of functional redundancy should nevertheless be evaluated within each ecosystem rather than assumed to be identical across environmental and host-associated microbiomes.

The opposite limitation also matters: similar species-level composition does not guarantee identical genomic function. Strains within the same species may differ in accessory genes, mobile elements, metabolic loci, or resistance determinants. Taxonomy can therefore reveal which organisms are associated with a phenotype, while direct metagenomics can reveal genomic features that cannot be recovered from taxonomic identity alone.

Predicted Function Is Not the Same as Direct Metagenomic Function

Functional prediction can be useful, but the analytical route must be described accurately. PICRUSt2 was developed to predict community functional potential from marker-gene sequencing profiles by placing observed sequences into a reference phylogenetic framework and inferring gene-family content from reference genomes (Douglas et al., 2020).

This approach can support exploratory pathway hypotheses when closely related reference genomes are available. However, prediction should not be interpreted as direct observation. Its accuracy depends on reference representation and evolutionary conservation, and it cannot fully capture previously unknown genes, strain-specific accessory genomes, horizontal gene transfer, or mobile genetic elements.

HUMAnN represents a different analytical concept. The method described by Franzosa et al. analyzes metagenomic or metatranscriptomic sequence reads directly, first identifying known species and mapping reads to species pangenomes, then analyzing previously unclassified reads to generate gene-family and pathway profiles (Franzosa et al., 2018). It should therefore not be grouped with marker-based functional prediction as though both tools infer pathways from a species abundance table alone.

Shotgun Depth Depends on the Biological Endpoint

There is no universal number of shotgun reads that separates "shallow," "deep," or "sufficient" metagenomics for every project. Required depth depends on community complexity, abundance distribution, host contamination, reference availability, library preparation, and the analytical output being requested.

Hillmann et al. demonstrated that shallow shotgun metagenomics can retain useful species-level taxonomic and broad functional information in suitable microbiome studies while requiring less sequencing than deeper whole-metagenome approaches (Hillmann et al., 2018). However, shallow sequencing is not equivalent to deep sequencing for all endpoints.

A 2026 benchmarking study by Treichel et al. systematically evaluated defined microbial communities across sequencing depths ranging from 0.1 to 50 Gb (Treichel et al., 2026). Under the tested conditions, relatively shallow data supported reference-based taxonomic profiling, while pathway-level information required greater depth and de novo MAG reconstruction required substantially deeper sequencing. The study also identified host DNA contamination and library preparation as important confounders in shallow metagenomics.

These findings support an endpoint-specific approach: a sequencing depth appropriate for relative community composition should not automatically be assumed sufficient for rare-gene detection, proteome-scale coverage, strain reconstruction, or MAG recovery.

Host-Rich Samples Change the Method-Selection Decision

Host DNA is particularly important when comparing taxonomy-focused methods with untargeted shotgun metagenomics. In shotgun sequencing, both host and microbial DNA enter the library. As the host fraction increases, a progressively smaller proportion of reads remains available for microbial analysis unless host depletion, enrichment, or substantially greater total sequencing is used.

Reduced-representation 2bRAD-M provides a different strategy for projects focused primarily on species-level community profiling. Sun et al. introduced 2bRAD-M as a species-resolved reduced metagenomic method and evaluated it with low-input, degraded, and host-contaminated DNA (Sun et al., 2022). Published experiments included total DNA inputs down to 1 pg and separate high-host-background simulations. These should be interpreted as experimental feasibility benchmarks rather than universal sample-input guarantees.

A later study specifically evaluated 2bRAD-M analysis for microbiome in host-rich samples without prior host depletion (Jiang et al., 2025). The study included mock samples containing more than 90% human DNA as well as saliva and oral tissue datasets. In the saliva comparison, 2bRAD-M captured host-specific and temporal community patterns similar to whole-metagenome sequencing while using approximately 5%–10% of the sequencing effort in that study. This result supports reduced metagenomic profiling as an option for selected host-rich, taxonomy-focused projects, but it should not be generalized into a fixed performance ratio for every sample type.

When deciding whether experimental host depletion is appropriate, see our guide on host DNA depletion requirements for microbiome sequencing. For difficult low-input samples, see how to obtain species-level microbiome profiles from low-biomass samples.

Practical Comparison: Taxonomy-Focused Profiling vs. Functional Metagenomics

For projects where shallow shotgun sequencing is appropriate, explore our shallow shotgun metagenome sequencing service. Projects requiring broader direct genomic coverage can instead use our metagenomic shotgun sequencing platform.

Decision Matrix: Match the Research Question to the Required Data

The following table provides an endpoint-first framework. It is intended for project planning rather than as a universal sequencing specification.

For broader technology comparisons, consult our overview of amplicon-based NGS vs. metagenomic shotgun sequencing.

How to Make the Final Method Decision

A defensible microbiome sequencing plan should answer five questions before sequencing begins:

• What is the primary statistical endpoint? Define the output table that will actually enter the downstream statistical analysis: species abundance, genes, pathways, transcripts, or genomes.
• Does the hypothesis depend on within-species genomic variation? If strain-specific genes or mobile elements matter, species identity alone is insufficient.
• How challenging is the sample? Low microbial biomass, DNA degradation, and high host background can alter the feasibility and efficiency of each sequencing strategy.
• How much functional evidence is required? Separate exploratory prediction from direct gene measurement, transcriptional activity, and metabolite measurement.
• What level of validation is needed? Candidate taxonomic or functional findings may require replication, orthogonal assays, or independent cohorts depending on their importance to the research conclusion.

This framework prevents both under-sequencing and unnecessary over-sequencing. A taxonomy-focused project should not automatically be converted into deep metagenomics simply because more genomic information is technically obtainable. Likewise, a study that depends on direct gene content should not substitute taxonomic prediction for the sequence evidence required to answer the hypothesis.

Summary: Choose the Minimum Data Type That Fully Answers the Research Question

Species-level taxonomic profiling and functional metagenomics are complementary tools rather than competing levels of quality. Species-resolved profiling is appropriate when the primary question concerns community composition, differential species abundance, ecological associations, or candidate taxonomic signatures. Direct shotgun metagenomics becomes more important when the project requires microbial genes, pathway components, mobile genetic elements, strain-specific content, or genome reconstruction.

• Choose species-level profiling when organism identity and abundance are the required primary outputs.
• Use predicted functional profiles cautiously for exploratory hypotheses, recognizing that prediction is reference-dependent and is not direct gene measurement.
• Choose direct shotgun metagenomics when genomic content itself is central to the research question.
• Choose deeper genome-resolved sequencing when MAG recovery or detailed genome reconstruction is required.
• Consider metatranscriptomics or metabolomics when the study requires evidence of transcriptional activity or biochemical state rather than genomic potential alone.
• Account for sample biology because host background, microbial biomass, and DNA quality can materially change the efficiency of each strategy.

The most efficient project is therefore not necessarily the one that generates the largest sequencing dataset. It is the one in which the biological hypothesis, sample properties, sequencing architecture, quality controls, and analytical deliverables are aligned from the beginning.

For projects involving difficult host-rich or low-biomass samples, explore our low-biomass and host-rich microbiome research solutions.

FAQ

References

1. 2bRAD-M reduced-representation metagenomics: Sun et al., Genome Biology 2022: Species-resolved sequencing of low-biomass or degraded microbiomes using 2bRAD-M.
2. Host-rich microbiome profiling without host depletion: Jiang et al., npj Biofilms and Microbiomes 2025: High-resolution microbiome analysis of host-rich samples using 2bRAD-M without host depletion.
3. Functional and taxonomic decoupling in microbial ecology: Louca et al., Science 2016: Decoupling function and taxonomy in the global ocean microbiome.
4. Predicted metagenome functions from marker-gene data: Douglas et al., Nature Biotechnology 2020: PICRUSt2 for prediction of metagenome functions.
5. Direct species-resolved functional profiling of sequence data: Franzosa et al., Nature Methods 2018: Species-level functional profiling of metagenomes and metatranscriptomes.
6. Shallow shotgun metagenomics evaluation: Hillmann et al., mSystems 2018: Evaluating the Information Content of Shallow Shotgun Metagenomics.
7. Sequencing-depth benchmarking: Treichel et al., Nature Microbiology 2026: Benchmarking of shotgun sequencing depth reveals the potential and limitations of shallow metagenomics and strain-level analysis.
8. 16S taxonomic resolution boundaries: Johnson et al., Nature Communications 2019: Evaluation of 16S rRNA gene sequencing for species and strain-level microbiome analysis.

For Research Use Only (RUO). Not for use in diagnostic procedures.