Soil shotgun metagenomics results can look intimidating because they combine millions of DNA reads with gene-family tables, functional pathway profiles and sometimes reconstructed microbial genomes. If you want to know how to interpret soil metagenomics gene abundance, functional annotation, pathway coverage, metagenome-assembled genomes or sequencing normalisation, start by identifying exactly which computational output you are reading.
Shotgun sequencing samples DNA fragments from many genomic regions rather than intentionally amplifying one marker such as 16S rRNA or ITS. That broader scope is useful for examining community genetic potential. It does not mean a report has measured live microbial cells, verified an enzymatic reaction or proved a complete ecological pathway is active. Read counts and annotations must pass through quality control, reference matching and explicit choices about how a sample is normalised.
This guide explains soil shotgun metagenomics interpretation step by step: field collection, host or plant DNA, read quality, taxonomic profiles, gene families, functional pathways, assembly, genome bins, coverage, annotation uncertainty, between-sample comparisons and plain-English reporting. Worked numbers are illustrative teaching examples, never represented as an actual field measurement. The goal is to translate computational output into statements a reader can check rather than into claims the data do not support.
Quick answer: what does shotgun metagenomics measure?
A shotgun metagenomic experiment sequences a sample of total recovered DNA fragments. Analysis software attempts to determine which organisms or genetic functions those sequences resemble. Depending on the workflow, outputs may include classified reads, gene-family abundance, gene-length-normalised abundance, estimated pathway coverage, contig assemblies and metagenome-assembled genome bins. These outputs answer related but different questions. None is a perfect census of the original soil.
The safest opening sentence specifies the evidence unit. For example: ‘The pipeline detected sequence reads assigned to genes associated with a metabolic pathway.’ That observation is different from saying ‘The pathway ran faster in this soil.’ Genes can be present without expression, protein activity or the necessary environmental conditions. Translating metagenomics means keeping capacity, expression and measured process rate separate throughout the explanation.
Why soil is a difficult substrate for genome-scale surveys
Soil contains mineral particles, humic substances, plant materials, fungal structures, microorganisms and extracellular DNA. A small aliquot samples only part of this spatially heterogeneous environment. Extraction efficiencies differ between organisms and soil types. Some cells are easy to break; others resist lysis. High organic content may complicate DNA purification. A comparison across two soil types therefore reflects both biological composition and what the protocol recovered.
A strong methods section identifies sampling depth, soil mass, collection and storage, homogenisation, extraction controls, DNA yield and independent field replication. Pooling several cores can provide an average view of a plot, but it does not create independent observations of each component. Treat extraction batch as a possible source of artefacts. Metagenomic software cannot recover organisms that were never sampled, DNA that was destroyed or reads systematically excluded before analysis.
Shotgun metagenomics versus amplicon sequencing
Amplicon sequencing targets a predefined genetic marker through PCR. Shotgun sequencing samples DNA from many genomic positions. Marker surveys can be efficient for community comparisons, while shotgun data can provide a wider range of taxonomic and functional information if depth, reference coverage and processing allow. Shotgun methods also capture DNA not belonging to the microorganisms of interest, including plant material and other environmental sources.
Neither method automatically dominates the other. If the question concerns shifts in detected marker communities, the soil ASV guide provides a targeted reading route. If the question concerns the repertoire of recoverable genes, shotgun metagenomics can be more relevant. A fair comparison also acknowledges cost, data complexity, genome size effects, sequencing depth, extraction bias and the practical limits of available references.
First decode the sample identifiers
A file named with a technical instrument identifier may contain little visible information about the collection site, depth or field replicate. The metadata table supplies those relationships. Without trustworthy sample matching, a functional difference may be assigned to the wrong treatment or date. Study design is therefore the first part of analysis, not an administrative afterthought appended to the final report.
Build a sample ledger with each biological unit, soil horizon, treatment, sampling date, extraction batch, library-preparation batch and sequencing run. Keep technical replicates distinct from independent field replicates. If several samples were pooled, record which original materials entered the pool. A sequencing result can be computationally flawless and scientifically misleading if its biological labels are wrong. Verification begins with identity before statistics.
Read the raw-read and quality-control summary
Sequencing machines produce reads with variable quality. Adapters, low-quality bases, duplicate fragments and sequencing artefacts can affect analysis. Quality-control pipelines trim or remove reads according to documented rules. In soil, a considerable share of recovered DNA may originate from plants or organisms outside the intended microbial comparison. Some workflows filter selected unwanted sequence sources; others analyse them as part of the sample.
Look for total reads, retained bases, read-length distribution, quality metrics, filtering rules and the fraction classified or excluded. A sample with half the usable sequence depth of another may detect fewer rare genes for technical reasons. A large percentage of unclassified reads is not necessarily useless: the soil environment is diverse and references are incomplete. Still, the exact definition of ‘unclassified’ must be recorded so it is not mistaken for missing data or contamination without evidence.
Contamination is not always visibly dramatic
Environmental metagenomes can include DNA introduced by collection tools, laboratory reagents, preparation equipment or cross-sample handling. Extraction blanks and processing controls can help identify these risks, especially where microbial biomass is low. A few reads matching a familiar laboratory-associated species do not establish contamination by themselves; context, controls and prevalence patterns matter. Similarly, removing all features found in any blank may discard legitimate sample evidence.
A useful report explains whether blanks were sequenced, which thresholds were used and whether samples were balanced across extraction and sequencing batches. If every sample in one treatment received one preparation procedure while controls received another, the two influences cannot be separated cleanly. Ask whether a surprising functional signal appears consistently in independent samples and whether it could arise from a processing artefact. A confident ecological label should be earned through the control structure.
Taxonomic classification is an assignment, not an organism count
A shotgun taxonomic profiler maps sequence evidence to reference organisms or marker sets. Algorithms differ: some classify all reads, some rely on taxon-specific markers, and some compare k-mer signatures. The resulting relative abundance can depend on database coverage, genome lengths, shared genes and normalisation choices. A label assigned to a genus or species is a computational interpretation, sometimes with ambiguity among close relatives.
Read the software, reference database version, minimum support settings and how unclassified material was handled. A pie chart of species estimates is not a direct microscopic census. Some organisms have no close reference genome; closely related organisms may share many genetic regions. Use phrasing such as ‘sequence evidence was classified as consistent with’ when resolution is uncertain. Scientific clarity does not require avoiding names, only avoiding stronger identity claims than the method supports.
What is a gene-family table?
A gene-family table groups reads or inferred gene sequences according to homologous gene categories. Depending on the system, rows may be functional orthologues, protein families, enzyme codes or named gene sets. Columns represent samples and values may be raw counts, length-normalised values, relative abundance or another scale. These are different statistical units even when all are casually called abundance.
Start by inspecting the legend and table headers. Ask whether gene families were assigned from raw reads, predicted coding sequences in assemblies, or a reference-guided mapping pipeline. If two related proteins are grouped into one family, the group may contain multiple biological roles. Conversely, one process may involve many distinct gene families. A table row is a computational category, not a one-to-one measurement of an ecological reaction.
Sequence similarity does not guarantee an exact function
A sequence can resemble a functionally characterised protein without performing precisely the same biochemical reaction. Closely related enzymes may accept different substrates or operate under different cellular conditions. Short reads sometimes cover only the conserved part of a gene, reducing specificity. Annotation pipelines use thresholds, reference models and naming conventions that can introduce uncertainty at several levels.
Interpret a predicted function as an evidence-supported hypothesis. If the story depends on a particular enzyme, check annotation specificity, gene length or coverage, and whether catalytic motifs or other functional evidence are available. Avoid promoting a broad gene-family match into a claim about an exact metabolite transformation. Independent validation, such as gene-context analysis, protein measurements or targeted activity assays, can strengthen the bridge from similarity to biochemical function.
Count tables often have hidden units
Suppose one sample contains 4,000 reads assigned to a gene family and another contains 6,000. If the first has 4 million quality-controlled reads and the second has 12 million, raw counts do not mean the family is more prevalent in the second sample. Sequencing effort differs. A crude reads-per-million comparison gives 1,000 and 500 family reads per million total reads respectively, even though the second raw count is larger.
That calculation is illustrative, not a recommended universal normalisation for gene length or mapping bias. The appropriate unit depends on the pipeline and biological question. A good caption states whether values represent raw fragments, reads per kilobase, estimated copies per million, proportions or something else. Never compare figures from two software tools solely because both call their axes ‘abundance.’ The hidden denominator is frequently the real explanation for apparently conflicting results.
Gene length influences read allocation
Long genes can attract more sequenced fragments simply because they occupy more DNA sequence. Some functional profilers divide read evidence by reference length to obtain a length-adjusted estimate. HUMAnN, for example, describes RPK-style units based on reads per kilobase before optional renormalisation. A reader must distinguish such units from raw reads and from the amount of gene-containing DNA per gram of soil.
Length correction addresses one source of bias, not every source. Coverage depends on taxonomic composition, extraction, sequence uniqueness, mapping thresholds, sequencing effort and reference completeness. If a method reports a family estimate corrected for gene length, say so. Do not treat a length-adjusted value as proof that the underlying enzyme was expressed more often, or that soil chemistry changed at the same rate.
Normalisation can change the ranking of samples
A sequencing table is often scaled so values are comparable across libraries of different sizes. Total-sum scaling converts family values into relative fractions or per-million equivalents; some pipelines use specialised procedures to account for composition and other complexities. The treatment of unmapped and unknown reads influences the sum that serves as the denominator. A change in that denominator can shift all reported proportions without a matching change in the biological sample.
Before writing ‘gene family X increased,’ identify the normalisation method and compare the same unit across samples. If one table includes unmapped material in its total and another excludes it, equal-looking numbers may not mean the same thing. For pipelines that use copies-per-million-like terminology, consult their documentation rather than assuming the abbreviation means raw counts per million. The word ‘copy’ may be a model-based normalised quantity, not a laboratory-measured absolute DNA count.
Stratified tables connect functions to tentative taxonomic sources
Some metagenomic outputs present an unstratified community total for a gene family alongside entries attributed to particular taxa. These stratified values attempt to allocate the family signal among recognised contributors. They can help test whether a potential function is broadly distributed or concentrated among particular lineages. However, assignment depends on available references and the ability to distinguish similar genes in related organisms.
If a family appears in the community total but not fully in named strata, do not assume the remainder is measurement error. Unclassified contributors and ambiguous mappings can be involved. Ask whether the report’s total includes all strata and unclassified components, or whether separate normalisation conventions apply. When translating the table, preserve the relation between community estimate and contributor estimates rather than adding numbers that were never designed to be summed directly.
Pathway abundance is not pathway activity
A pathway is a model of biochemical steps associated with a biological transformation. Software may estimate its abundance by combining evidence from relevant gene families under a reference pathway definition. The estimate describes genetic potential inferred from the DNA, not the observed flux of metabolites through the pathway. Real processes depend on expression, enzyme activity, temperature, moisture, substrates, oxygen and other constraints.
A thoughtful statement might read, ‘Sequence annotations supported a greater relative representation of genes associated with this pathway.’ The tempting alternative, ‘The soil performed this pathway faster,’ needs process measurements. A pathway’s genes can be present in organisms that are dormant, damaged or prevented from operating by local conditions. Soil incubation assays, enzyme measurements, gas fluxes or isotopic tracers ask complementary questions about actual processes.
Pathway coverage and pathway abundance are not synonyms
Coverage may express the fraction of expected pathway components supported by sequence evidence, whereas abundance estimates the relative representation of an inferred pathway under an algorithm. A pathway can show high coverage but modest abundance, or strong evidence for several abundant components while failing a completeness threshold. Definitions depend on software and databases, so a number labelled ‘0.8’ needs its actual metric.
Do not translate an 80% coverage estimate into ‘80% of the metabolic pathway occurred.’ A pathway is not a progress bar for completed chemistry. An incomplete annotation may arise from reference gaps, shallow sequencing, alternative enzymes or genuinely missing steps. Preserve the algorithm’s pathway model and threshold when communicating the result. Where function matters, compare alternative models and independent biological evidence before giving a mechanistic conclusion.
Assembly joins overlapping fragments into contigs
Instead of classifying each short read separately, an assembler may use overlap or graph-based relationships to reconstruct longer contiguous sequences called contigs. In diverse soil communities, close strains, repeats, uneven genome coverage and low-abundance organisms make reconstruction difficult. Assembly can improve access to gene neighbourhoods but may omit important material or accidentally join ambiguous regions if the evidence is insufficient.
Read the total assembly length, contig-length distribution, N50 where reported, minimum contig cutoff and the fraction of reads mapping back to the assembly. N50 is a summary of contig lengths, not a certificate that all reconstructed sequences are correct. A large assembly is not automatically biologically more complete than a smaller one. Consider coverage, error checks and study design before comparing assemblies from different soil environments.
Mapping reads back to contigs estimates support
After assembly, reads may be mapped to contigs to estimate depth or coverage. Average depth reflects how many read bases cover each reference position under mapping rules. Breadth describes how much of a contig has coverage. These quantities offer clues about representation and reconstruction reliability, but they are not direct counts of living cells. Shared sequences and strain variation can make mapping ambiguous.
Suppose a contig is covered across only a small section at high depth. Reporting ‘high coverage’ without separating breadth from depth could mislead. Examine uneven coverage, suspect regions and multimapping policies when a specific gene claim depends on a contig. A defensible annotation states which sequence region supports it. High depth over a short conserved segment is weaker evidence for a complete novel gene than broad, consistent coverage.
What does a metagenome-assembled genome mean?
A metagenome-assembled genome, commonly called a MAG, is a set of contigs computationally grouped as likely originating from one population genome. Binning algorithms may use sequence composition, coverage patterns and other signals. A MAG is a reconstruction from environmental DNA, not necessarily a cultivated organism or a fully closed chromosome. Complex soil samples can yield incomplete or mixed bins.
Inspect estimates of completeness and contamination, how they were computed, assembly and binning methods, and whether the genome was dereplicated against similar bins. Marker-gene quality estimates have assumptions and may be unreliable for unusual lineages. A high-quality classification should be explained through the applicable standard and data, not asserted from an attractive completeness percentage alone. Treat MAG-derived functional predictions as properties of a reconstructed sequence set under those limits.
Completeness and contamination estimates have boundaries
Genome-bin quality tools commonly compare expected marker genes and their apparent multiplicity. Missing markers can suggest incompleteness; duplicated markers can suggest contamination or strain mixture. However, real biological variation and atypical genomes can complicate the inference. A headline such as ‘95% complete, 2% contamination’ is a model-based estimate whose meaning depends on the reference expectations and workflow.
When communicating such numbers, identify the tool and avoid calling the MAG 95% of a proven individual organism. A reconstructed bin may contain errors that marker metrics do not detect. Independent read mapping, taxonomic consistency and other quality checks provide supporting evidence. If a proposed ecological function is encoded on one questionable contig, its association with the bin deserves extra scrutiny even if the overall summary metrics look impressive.
A MAG is not an isolate and not a field organism
Culturing yields a physical isolate that can be tested under controlled conditions. A MAG yields a genomic reconstruction from environmental sequence evidence. It can suggest metabolic capabilities and evolutionary relationships, yet it does not demonstrate a live isolate with the predicted phenotype. Different strains may share enough sequence to complicate binning, while highly similar functions can occur in organisms with very different ecological behaviour.
Write ‘the reconstructed genome encoded genes annotated as’ instead of ‘the organism carried out.’ This small linguistic adjustment preserves the observation. The distinction is especially important in claims about carbon storage, nitrogen cycling or pollutants: genes support hypotheses about potential transformations, while activity requires environmental measurements or targeted laboratory validation. A useful MAG opens new questions rather than closing every biological uncertainty.
Relative abundance, absolute genes and soil mass
Most shotgun abundance outputs are conditioned on the recovered DNA and sequencing library. They do not automatically convert to copies of a gene per gram of dry soil. Such conversion requires a defensible quantitative standard, sometimes involving spike-ins or complementary measurement, as well as assumptions about extraction recovery, genome sizes and sample masses. The method should explain every conversion factor.
A paired soil qPCR assay can quantify selected genetic targets under its own constraints. Shotgun metagenomics can supply broader potential-function profiles. If both show compatible changes, confidence may improve, but they need not yield interchangeable quantities. They may target different sequences and normalise against different denominators. In prose, label ‘relative shotgun signal’ separately from ‘qPCR-estimated copies per gram.’
Metagenomics versus metatranscriptomics
Metagenomics focuses on DNA and therefore community genetic potential. Metatranscriptomics sequences RNA-derived material to investigate transcripts detected under particular conditions. RNA measurements can provide clues about expressed genes at sampling time, but transcript abundance is not an automatic measure of protein abundance or reaction rate. RNA preservation, extraction, depletion and normalisation create their own uncertainties.
A gene detected in DNA but not RNA does not prove permanent inactivity; the relevant expression could be low, transient, localised or missed by sampling. Likewise, a transcript detection supports expression-related evidence without measuring enzyme throughput. A convincing process explanation often combines DNA, RNA, proteins, metabolites and physical conditions. Each method contributes a separate link in a causal chain.
Read the environmental measurements alongside gene profiles
A functional gene category becomes more meaningful when paired with pH, soil moisture, oxygen availability, temperature, substrate concentration or measured nutrient pools. For instance, genes associated with anaerobic processes may be present in aerated soil without that pathway dominating current activity. Environmental measurements help identify which predicted functions are plausible under observed conditions.
Do not treat environmental correlations as proof of causality. Sites may differ simultaneously in management, mineralogy, plant cover and sampling depth. Use well-planned experiments or other causal evidence when making treatment claims. Link the explanatory chain carefully: available genes, relevant environmental setting, observed transcripts or proteins, measured flux and ecological outcome. A missing link is an open research question, not an invitation to fill it with certainty.
Soil enzyme assays and the question of function
An enzyme assay measures a specific laboratory-defined catalytic potential under chosen substrate, pH, temperature and incubation conditions. A metagenomic family annotation describes DNA evidence associated with a possible function. These methods may agree, but their units, biases and biological meaning are different. A rise in an enzyme-associated gene family does not guarantee the corresponding assay rate will rise proportionally.
Use the soil enzyme activity guide when comparing measured reaction signals with sequence predictions. Explain whether the enzyme assay uses artificial substrate, reports units per dry mass and includes background correction. When outcomes differ, consider regulation, protein persistence, environmental limitation, alternative pathways and assay selectivity. This is a productive scientific conversation, not an embarrassment to hide.
An illustrative pathway example with two soils
Imagine a study reporting that a reference-defined pathway has 12 normalised units in soil A and 18 in soil B. A naive claim is that the relevant reaction occurred 50% faster in B. The ratio 18 divided by 12 is 1.5, but it applies only to the reported normalised pathway measure. It does not identify an absolute biological flux unless that measure was explicitly calibrated and validated against activity.
A responsible explanation is: ‘The pathway’s estimated representation in the selected metagenomic profile was 50% higher in B on this scale.’ Follow with the units and normalisation convention. Then examine independent replicates and uncertainty. If respiration or isotope-based process rates are also available, discuss them separately before relating them to the genomic potential. This preserves both the numerical calculation and the inferential boundary.
Watch the unmapped and unknown fraction
A meaningful share of environmental sequence reads may fail to map to available reference genes or genomes. Soil is especially diverse, and novel lineages and proteins can be poorly represented by curated libraries. An ‘unmapped’ category may mean no acceptable reference hit under particular parameters; it does not mean the reads contain no biological information. An ‘unknown function’ label may reflect limits in current annotation systems.
If a study reports that the proportion of mapped reads rises after treatment, ask whether sequencing quality, sample composition or database assumptions changed. The increase might represent a genuine biological change, or it might reflect a better match between recovered DNA and known references. The fraction not explained by a model deserves visibility. Hiding it can make an inferred functional profile appear more complete than it is.
Comparative metagenomics requires design-aware statistics
Two samples with different functional profiles may reflect field treatment, geography, soil depth, plant roots, seasonal conditions or differences in sequencing library preparation. A statistical analysis should represent independent experimental units and account for relevant covariates where possible. Many gene families and pathways may be tested, so multiple-testing control and effect-size reporting are important. Standard assumptions can be challenged by sparse, compositional and highly correlated features.
Inspect how zeros, low-abundance functions and different sequencing depths were handled. A bar plot of group means can hide one outlying sample. Request or inspect individual replicate values and confidence intervals when they exist. Significance does not prove a causal mechanism, nor does a nonsignificant result establish identity of communities. The analytical goal is to distinguish plausible stable patterns from artefacts, uncertainty and chance.
Gene categories can overlap and cannot always be added
Some databases represent pathways hierarchically, while genes participate in several processes. Adding the abundances of overlapping categories can double-count the same sequence evidence. A gene may have more than one plausible annotation, and a pathway algorithm may combine family signals using non-additive rules. The plain-English summary should not assume every listed line represents an independent reservoir of genetic capacity.
Study the output format and official software documentation before calculating totals. Ask whether values are already normalised, whether the categories are mutually exclusive and whether community or stratified levels are being mixed. One table may be designed for ranking functions, another for estimating a particular pathway within a model. Calculations should follow the data model rather than the intuition that any column of numbers can simply be summed.
Functional interpretation and microbial interactions
Metagenomics can suggest which genomic capabilities coexist within a sample, but it does not establish that two organisms interact directly. A pathway may be distributed among different taxa, or several alternative routes may lead to similar products. Short-read taxonomic assignment may not link every functional gene confidently to its source organism. A reconstructed genome may improve that link but can still be incomplete or contaminated.
If you propose a metabolic handoff between organisms, label it as a hypothesis unless gene localisation, metabolite exchange or experimental evidence supports the interaction. Microbial ecology often involves spatial microhabitats invisible to bulk soil sequencing. Two genes detected in the same composite sample need not occur in neighbouring cells. Preserve scale: shared sample, shared reconstructed genome, same cell and demonstrated interaction are different claims.
Why time and disturbance matter
A soil community can change rapidly after moisture pulses, temperature shifts, substrate additions or physical disturbance. DNA may persist after activity declines, whereas RNA, proteins and metabolites have different response times. Comparing a sample taken immediately after rewetting with one taken weeks later can produce differences unrelated to an underlying long-term management effect unless the time frame is controlled.
Document collection timing, recent weather, storage delay and disturbance history. A single snapshot is valuable for describing genetic potential at a defined moment, not for establishing permanent pathway activity. Longitudinal designs can reveal trajectories, but repeated samples from the same location require appropriately paired analysis. Translating results into advice for land management should reflect the timescale actually measured and avoid universal claims from one sampling event.
Communication case: fixing a misleading headline
Suppose a report says ‘Soil metagenomics proves carbon fixation doubled.’ You inspect the data and discover a 2.0-fold rise in the relative abundance of a small set of genes annotated to a carbon-fixation pathway. No direct carbon uptake rate, isotope tracing or expression evidence was provided. The headline equates potential with realised flux and omits the compositional denominator.
A more defensible line is: ‘The analysed libraries showed approximately twice the normalised relative signal for genes annotated to the specified carbon-fixation pathway, suggesting a hypothesis for further process measurements.’ The second version is precise without being inaccessible. It makes a concrete claim that could be rechecked and identifies the follow-up evidence needed. Good technical writing preserves the observable result and the uncertainty at the same time.
A compact protocol for reading a new metagenomic report
First write down the biological question and sampling unit. Second identify how DNA was obtained, filtered and sequenced. Third name the output type: read classification, gene families, pathway model, contigs or MAGs. Fourth record the exact unit and normalisation rule. Fifth inspect controls, unclassified material, replicate structure and uncertainty. Only then describe the pattern in plain language.
This sequence prevents numbers from migrating between incompatible analytical levels. A genome completeness estimate should not be described as a percentage of soil organisms recovered. A pathway coverage score should not become an enzyme-rate measurement. A copy-per-million-like normalisation value should not become copies per gram of soil without an independent bridge. The protocol works because it forces measurement and inference to remain visible as separate stages.
Vocabulary that prevents misinterpretation
In scientific English, ‘potential’ describes capability indicated by evidence, not necessarily realised performance. ‘Annotation’ means a computational assignment, not a guaranteed biochemical function. ‘Coverage’ may describe read mapping or pathway component support, depending on context. ‘Abundance’ is incomplete without a unit. ‘Complete genome’ should not be conflated with an incomplete genome bin, and ‘active’ requires a separately justified criterion.
When translating metagenomics for students, policymakers or another language community, define these terms at first use. The cross-linguistic transfer guide explains why near-equivalent terms can mislead; translation failures involving context and meaning show why preserving relationships between ideas matters more than copying individual words. Treat units and uncertainty as part of the vocabulary lesson.
Frequently asked questions
Does shotgun sequencing find every soil gene? No. Coverage depends on recovered DNA, sequencing depth, reference databases and analysis, and rare genomes can remain underrepresented. Is a gene-family table the same as an enzyme assay? No; it reflects DNA evidence for annotated categories, not measured catalytic activity. Does a MAG represent a cultured organism? No; it is a reconstructed genomic bin from environmental sequence evidence. Is ‘unknown’ the same as contamination? No; insufficient reference matching is one possibility.
Can two software packages produce different pathway values? Yes, because databases, models, mapping and normalisation vary. Is high pathway coverage proof that every chemical step occurred? No; the output describes evidence for predicted components. Can metagenomics prove a treatment improved soil health? Not by itself; the question needs ecological objectives, study design and complementary outcomes. Is raw read count a fair abundance comparison across libraries? Not without considering depth and other biases.
Practice questions with explained answers
Sample A has 1,500 reads assigned to a selected gene family from 3 million retained reads. Sample B has 2,000 from 8 million. A naive count ranks B higher. Normalising just by retained-read totals, A has 500 reads per million and B has 250, so A has the larger share on that simplified scale. The exercise demonstrates why denominators can reverse a ranking, not why reads-per-million always solves metagenomic compositional problems.
Now suppose a pathway abundance estimate rises while a measured soil respiration rate remains unchanged. Is one necessarily wrong? No. DNA-level genetic potential, expression, available substrate and current process rates are distinct. Next suppose a MAG contains a putative denitrification gene. Does that prove denitrification occurred in situ? No. It supports an annotation about a reconstructed genome, requiring environmental and experimental evidence to establish activity.
Further learning and authoritative sources
Use the EMBL-EBI MGnify overview and MGnify training introduction to understand public microbiome data and analysis types. The HUMAnN documentation explains gene-family and pathway units, including length adjustment and optional normalisation. Method descriptions are necessary for interpreting outputs; they do not validate claims made without suitable controls and study design.
Within eduKateSG, connect this guide to soil ASV sequencing, soil qPCR, soil microbial biomass and soil respiration. These methods describe different dimensions of soil biology, so comparison is most useful when the units and assumptions remain separate.
Closing principle: DNA potential is a starting point
Shotgun metagenomics reveals an extraordinary amount of information about DNA recovered from complex microbial communities. Its outputs can help identify likely taxa, predicted gene families, pathway models and reconstructed genomes. But each result is a translation from physical material through instruments and software into a biological claim. The last stage requires judgement about what was actually measured.
When reading a report, keep asking: what is the denominator, what is the reference, what is the confidence, and what additional experiment would turn a hypothesis into a measured process? The answers make metagenomics more useful, not less impressive. They allow a student or professional to communicate real biological insights without converting sequences into imaginary rates, organisms or certainty.