A soil metagenome-assembled genome, or MAG, is a reconstructed draft genome assembled from the DNA of a mixed microbial community. When you want to read MAG quality reports, CheckM or CheckM2 completeness estimates, contamination percentages, genome coverage tables or genome taxonomy assignments, start by asking how the DNA fragments were assembled, which contigs were grouped together and how the quality numbers were estimated.
A MAG is not automatically the complete genome of one verified living organism. Genome reconstruction is a computational inference from sequences in a mixed sample. Two related strains can be confused, rare genomes may be underrepresented, and a bin can accidentally contain contigs from more than one biological source. Completeness and contamination are estimates, not ground truth, and a high-looking score cannot guarantee every metabolic gene is present.
This guide explains the steps needed to interpret soil MAG reports accurately: contigs, coverage, binning, quality scores, taxonomic placement, dereplication, functional annotation, uncertainty and practical validation. Worked numerical examples and diagnostic questions show how to translate a genome-bin spreadsheet into precise educational language without claiming that a draft assembly is a cultivated isolate or an observed ecological activity rate.
Quick answer: what does a soil MAG report tell us?
A metagenome-assembled genome is a collection of assembled DNA sequences inferred to originate largely from the same organism or closely related population. A report commonly lists a bin identifier, assembled length, number of contigs, estimated completeness, estimated contamination, sequence coverage and a proposed bacterial or archaeal taxonomy. It may also include predicted gene counts and reconstructed metabolic pathways. The genome is a useful model for a microbial population represented in the sequenced material, not a photograph of one cell.
Each field addresses a different risk. Completeness asks how much expected genomic content may have been recovered. Contamination asks whether extraneous genomic material may have entered the bin. Fragmentation describes how many pieces the genome still occupies. Coverage expresses how sequencing reads map back to the assembly. Taxonomy identifies evolutionary relationships using a chosen reference and computational framework. Read these fields together; none can certify an assembly alone.
Why genome-resolved metagenomics differs from functional gene profiling
A conventional shotgun metagenomic table may list all DNA fragments assigned to particular protein families across a mixed community. A MAG analysis attempts to organise fragments into provisional genomes, potentially connecting genes that appear to belong to the same organism. This permits hypotheses about which pathways can coexist within one population genome and can help classify microbes lacking cultivated representatives. Yet successful binning depends on adequate assembly, coverage and distinguishable sequence patterns.
The distinction matters when a scientist claims one microorganism can carry out a whole reaction sequence. A community functional table may contain the relevant genes distributed among different organisms; a well-supported genome bin can provide stronger evidence that those genes co-occur within the same reconstructed genomic population. Even then, the result is genetic potential, not measured pathway operation. See the soil shotgun functional-gene guide for the community-table perspective.
Start with the sample and assembly method
Genome recovery begins with actual soil material, DNA extraction and sequencing. The sample may come from bulk soil, root-associated material or a particular depth and season. Those conditions determine which microorganisms are represented in the DNA and which are missed. The library may contain plant, fungal, bacterial, archaeal and other DNA in very different proportions. Genome assembly then depends on the quantity and quality of reads assigned to each population.
Ask whether the assembly was built from each sample independently or from pooled samples. Co-assembly may improve sequence recovery for organisms shared across samples, but it may also complicate resolution of similar strains and the interpretation of sample-specific abundance. The assembly program and parameters should be documented. A MAG identifier without its input sampling and assembly history is not sufficient provenance for a confident ecological claim.
Reads overlap to form contigs
A sequencing read is a relatively short stretch of DNA observations. Assembly algorithms use patterns of overlap or related graph structures to reconstruct longer consensus sequences called contigs. In a clean single-organism dataset, this can sometimes produce long pieces that reflect much of a genome. In soil, many related organisms contribute similar sequences, repeats complicate reconstruction, and some populations receive uneven sequencing depth. The resulting contigs are often fragments rather than complete chromosomes.
Contig length and number reveal part of the assembly’s structure. A genome bin containing 200 contigs is much more fragmented than one containing five long contigs of comparable total size, although fragmentation alone does not determine correctness. Short contigs may not carry enough sequence context for confident taxonomic assignment or functional reconstruction. Conversely, a long contig may still contain a misassembly. Treat contig continuity as one quality dimension, not as proof of biological identity.
A bin is a proposed collection of related contigs
Binning algorithms group assembled contigs using signals such as sequence composition, coverage patterns across samples and relationships to known reference sequences. Contigs from the same genome often have related characteristics, but the signals overlap among organisms. Closely related strains may share much of their genomic composition and coverage. A contig encoding a common gene may resemble sequences from multiple populations and resist unambiguous placement.
The label ‘MAG-12’ is therefore a working identifier, not a newly discovered species name. It identifies a set of sequences created by a particular binning and refinement workflow. If the workflow changes, the same original reads could produce a different group of contigs and a differently numbered bin. Keep the assembly file and bin membership information whenever possible. The report should allow readers to trace which sequences were included, excluded or reclassified.
Read mapping helps reveal bin support
After assembly, original reads can be aligned back to contigs. Coverage usually summarises how many sequencing fragments overlap each genomic position, sometimes as a mean depth. Uneven coverage can suggest repeats, ambiguous mapping, mixed strains, assembly errors or genuine genomic features such as varying copy number. Coverage should be interpreted with the aligner’s settings and the treatment of multimapping reads in mind.
A coverage of 20 times does not mean the organism was twenty times more abundant than another organism. It means that, under a particular calculation, positions in the assembled sequence were covered by reads at a certain average depth. Differences in genome length, extraction, sequencing depth, reference bias and cross-mapping may affect comparisons. For ecological abundance, ask which reference set, normalisation and sample-level model were used rather than treating coverage as an absolute cell count.
Worked example: calculate approximate average coverage
Imagine a classroom example with a 2-million-base draft assembly and 40 million mapped read bases that contribute uniquely to that bin under the chosen filtering rule. Dividing 40 million bases by 2 million assembled bases gives an average depth of 20-fold coverage. This arithmetic assumes each mapped base contributes to the coverage numerator as expected and ignores complex alignment issues. It is a starting interpretation, not a replacement for per-contig coverage inspection.
Now imagine that half of those mapped bases concentrate in a small repeated region while other regions receive little or no coverage. The same crude mean can conceal weakly supported sections. Look for a depth distribution or coverage plots, especially around suspicious contig junctions. A reasonable MAG report may provide several coverage summaries because the average alone can be misleading. The meaning of any ’20×’ label depends on how reads were filtered and counted.
Differential coverage can separate populations
When the same environmental assembly is profiled across multiple samples, contigs from the same population may change in abundance together. Binning tools can use these joint coverage patterns to help distinguish genomic sources. If two contigs have similar nucleotide composition and highly correlated coverage across many independent samples, grouping them may become more plausible. Yet correlated coverage does not uniquely establish common genomic origin; species can respond similarly to the same environment.
Sampling design affects this signal. When many samples share the same treatment gradient, ecologically co-occurring organisms may appear correlated and complicate binning. If there are only one or two samples, differential coverage provides limited discrimination. A good interpretation describes coverage as one line of support combined with sequence composition and other evidence. Resist the temptation to turn a compelling scatter plot of coverage into definitive proof of a single genome.
Why closely related strains complicate soil assemblies
Different strains of a species can share many genes while carrying divergent accessory sequences and distinct alleles. A mixed soil sample may contain several related strains at once. Assembly can collapse highly similar regions into consensus contigs and separate variable segments, producing bins that do not correspond perfectly to any individual cell’s genome. Binning and quality estimates become particularly challenging when these strains overlap in abundance and sequence composition.
A recovered MAG may represent a population consensus rather than a pure strain genome. Describing it as ‘the genome of this exact individual strain’ can overstate the resolution. A useful report may discuss strain heterogeneity, microdiversity, single-nucleotide variants or assembly graph complexity. If these were not assessed, state the more limited claim: the bin contains genomic sequences associated with a microbial population. Detailed strain conclusions need additional evidence and methods.
Completeness is an estimate of expected genomic content
MAG completeness is often estimated using sets of marker genes expected to occur in most genomes of a relevant lineage, often as single copies. If many expected markers are present, the bin may contain a large fraction of a typical genome’s required content. The method assumes the marker set is suitable for the lineage and that the detected sequences are correctly predicted and assigned. No software possesses direct access to the uncultivated organism’s missing genome.
This means a reported 95% completeness does not literally demonstrate that exactly 95% of every nucleotide in the real genome has been recovered. It is a model-based estimate. Organisms with unusual, reduced or highly divergent genomes may violate assumptions used to estimate completeness. A robust explanation gives the estimate and method together. It also avoids interpreting the missing five percent as a known list of absent ecological functions.
Marker genes provide a yardstick, not the genome itself
Think of a library where most books in a familiar series contain a predictable set of chapter headings. If a recovered bundle contains nearly all those headings, we might infer that most of the book is present. But an unusual edition could omit chapters legitimately, and duplicated headings could reflect a second book mixed into the bundle. Single-copy marker approaches make a more sophisticated version of this inference using genomic references and evolutionary assumptions.
The analogy reveals why lineage matters. A marker typical of many bacteria may not behave similarly in some archaea or highly reduced genomes. Check whether the quality tool adjusts expectations using a lineage-specific model or a more general predictor. When high completeness appears for an unusual organism, confidence also depends on assembly support, contamination estimates and independent quality indicators. The percentage is a summary of evidence, not an oracle.
Contamination is not simply dirty laboratory equipment
In a MAG report, contamination commonly refers to evidence that a bin contains genetic material from more than one source, or excess marker representation inconsistent with a clean single-population genome under the model. This can happen during computational binning when contigs from different populations are grouped. Laboratory contamination is another possibility but should not be assumed from the metric alone. Strain heterogeneity can also complicate marker-duplication signals.
If a quality report states 4% contamination, it does not necessarily mean that 4% of all sequenced bases were introduced by an unclean bench. The number is an algorithmic estimate of genomic impurity. To investigate, inspect coverage differences, taxonomic conflicts, contig composition, duplicated markers and assembly provenance. Clear terminology prevents a technical estimate from being translated into an accusation about laboratory practice unsupported by the evidence.
Why duplicate markers can suggest a mixed bin
Many quality tools use markers expected to occur as single copies. If a bin includes two distinct versions of a supposedly single-copy gene, one hypothesis is that material from two genomes has been combined. Other possibilities include legitimate genomic duplication, strain variation, misprediction or assembly artefacts. The software’s contamination metric depends on its model for interpreting such evidence across a marker set.
A single duplicated marker is not enough to declare the whole bin invalid. Check the number and distribution of suspicious markers and whether the contigs carrying them differ in coverage or taxonomy. Some organisms have genuine deviations from assumed marker copy counts. An informed reader treats duplicate-marker evidence as a diagnostic clue and considers alternative explanations rather than accepting the numerical estimate without context.
MIMAG supplies reporting categories, but read their conditions
The Minimum Information about a Metagenome-Assembled Genome, or MIMAG, framework sets expectations for reporting draft bacterial and archaeal genomes. Its medium-quality criteria include estimated completeness of at least 50% and contamination below 10%. High-quality draft criteria include more than 90% completeness and less than 5% contamination, together with additional requirements involving ribosomal RNA genes and transfer RNAs. The extra requirements are often omitted in casual descriptions.
A bin with 94% completeness and 2% contamination does not automatically satisfy every element of the full high-quality MIMAG designation if required RNA-gene evidence is unavailable. Likewise, terms such as ‘near complete’ and ‘high quality’ can be used under tool-specific schemes, so consult the actual reporting definition. The MIMAG publication is the appropriate primary reference for those categories. Use its criteria explicitly, not as a loose marketing label.
Worked example: classifying three hypothetical genome bins
Suppose bin A has 96% estimated completeness and 3% contamination, bin B has 72% completeness and 7% contamination, and bin C has 38% completeness and 2% contamination. Under the basic numerical thresholds, A meets the completeness and contamination part of the high-quality draft rule, B meets the medium-quality thresholds, and C falls below the medium-quality completeness requirement. This is a simplified classroom comparison, not a full MIMAG certification.
To determine whether A meets the full high-quality designation, check its RNA-gene and other required information. To interpret C, ask which genes and regions are missing rather than assuming it is worthless. A low-completeness bin can still provide useful sequence evidence for particular narrow questions, but broad claims about metabolic absence would be fragile. Always distinguish the quality category, the intended analysis and the exact evidence needed for a biological claim.
Completeness and contamination do not cancel each other
Some projects rank bins using a composite score, for example completeness minus a weighted penalty for contamination. Such a formula can be useful for prioritising manual review, but it does not erase the different biological consequences of missing content and mixed content. A contaminated bin may falsely suggest that genes from two organisms coexist. An incomplete bin may miss real pathways and incorrectly imply genetic incapacity. These are distinct risks.
A bin with 98% completeness and 15% contamination should not automatically be preferred over a cleaner, moderately complete bin merely because its numerical difference is favourable under one arbitrary score. The proper choice depends on the research question, the confidence of the estimates and specific suspected errors. If you seek one gene’s genomic neighbourhood, contamination and misbinning may be decisive. If you compare broad conserved markers, different limitations may matter more.
Assembly size is not enough to measure recovery
A MAG table often lists total assembled base pairs. A 3-megabase bin might be nearly complete for one organism and highly incomplete for another. Microbial genome sizes differ, and some symbiotic lineages have very small genomes. There is no universal genome-length threshold that establishes completeness. This is why MIMAG recommends reporting assembly statistics while relying on additional evidence for quality evaluation.
A larger bin can result from better recovery, contamination, duplicated regions or genuine biological genome size. Do not praise a longer assembly until you inspect its composition and quality. For comparisons across bins, include estimated completeness, contamination, contig counts and taxonomy rather than treating megabases as a biological ranking. Assembly size is a descriptive statistic, not a stand-alone certificate of genome quality.
N50 measures continuity, not correctness
N50 is an assembly continuity statistic. Arrange contigs from longest to shortest and find the length of the contig at which cumulative assembled length reaches at least half of the total. That contig length is the N50. A higher N50 generally indicates a larger share of bases in longer contigs, but it cannot determine whether those contigs are correctly assembled or properly assigned to the bin.
Suppose a simplified assembly has contigs of 400, 300, 200 and 100 kilobases, totalling one megabase. Half the total is 500 kilobases. The first contig supplies 400, and adding the second reaches 700, so N50 is 300 kilobases. The arithmetic is straightforward; the interpretation is limited. Two bins can share an N50 while differing drastically in completeness, contamination and support from mapped reads.
L50 and contig count add useful context
L50 is the number of longest contigs needed to account for at least half the assembled sequence length. In the preceding 400, 300, 200 and 100 kilobase example, two contigs reach more than half of the genome, so L50 is two. N50 answers a length question while L50 answers a count question. Neither explains whether the sequence belongs to one organism. Report them as complementary continuity measures when available.
A genome bin may have dozens of fragments but retain important diagnostic genes on well-supported contigs. Another may have a few strikingly long sequences created by a flawed join. If two assemblies are compared, ask whether they represent the same biological population and were built with similar technologies and filters. Continuity statistics are most informative when accompanied by maps of genome content, read support and the quality-estimation method.
GC content can help reveal inconsistent contigs
GC content is the fraction of guanine and cytosine bases in a DNA sequence. Genomes can differ in their typical GC composition, so an outlying contig may deserve inspection. Binning software may use nucleotide composition patterns, including tetranucleotide frequencies, to group or flag contigs. Yet within-genome variation can arise through mobile genetic elements, evolutionary history and different genomic regions. GC outliers are clues, not automatic contamination.
Imagine a bin whose contigs mostly cluster near one GC fraction while several long contigs have very different coverage and nucleotide composition. This combination strengthens the case for reviewing those contigs. But a genuine plasmid or horizontally acquired element could also differ from the chromosomal background. A careful interpretation combines taxonomy, coverage, composition and assembly links rather than setting a universal GC cutoff.
Plasmids and mobile elements complicate genome boundaries
A microbe can carry extra-chromosomal plasmids, and mobile genetic elements may move between lineages. These sequences can have distinct copy numbers, coverage and nucleotide composition relative to a chromosome. A binning workflow may exclude them, attach them incorrectly or group them with different organisms. Consequently, a MAG’s gene repertoire may not capture every mobile element in the environmental population, and a plasmid-like contig does not always belong unambiguously to one recovered bin.
This matters when a study claims that a particular taxon possesses a resistance or metabolic gene located on a mobile element. Short-read co-occurrence is not always enough to prove physical linkage to the proposed host. Long-read sequencing, read-pair connections, assembly graph evidence or independent validation may strengthen the assignment. State whether the gene is on a confident chromosomal contig, an uncertain accessory contig or simply detected in the broader sample.
Binning refinement can improve quality but also change the result
Many workflows combine outputs from several binning algorithms and use refinement steps to select or merge contig groups. Manual inspection can remove suspicious contigs or recover missed fragments. Improvements may increase estimated completeness, reduce contamination or alter taxonomic classification. But each change is a hypothesis about sequence ownership and should be documented. An attractive quality score after extensive optimisation is not necessarily independent confirmation.
A reproducible report ideally records the binning tools, versions, selected parameters and refinement criteria. If bins were merged, ask whether the merged contigs share consistent coverage and genomic markers. If contigs were removed, note whether functionally interesting genes disappeared. A reader comparing publications should avoid treating a bin identifier as immutable when a later study has reassembled or refined the underlying reads.
Taxonomic assignment uses evolutionary evidence, not just one name
Recovered genomes can be placed using conserved marker genes and comparisons against genome reference collections. Genome-based systems such as the Genome Taxonomy Database, or GTDB, provide a standardised framework that can include uncultivated lineages. A reported species or genus name depends on the reference release, phylogenetic methods and quality of the query genome. Poorly represented or highly divergent lineages may receive provisional labels rather than familiar names.
Check the classification tool and database release. If a genome is assigned only to a family or novel cluster, do not invent a species identity for reader convenience. A lineage may also be labelled differently under traditional nomenclature and a genome-based classification. The GTDB information page explains the reference framework. The appropriate conclusion names the assigned rank and acknowledges unresolved identity.
Average nucleotide identity answers a different question
Average nucleotide identity, or ANI, compares homologous genomic sequences between genomes and is widely used in species-level genome comparisons under defined thresholds and coverage requirements. It is not the same as a percentage of aligned reads in a metagenomic sample. Two bins can show high ANI over their shared regions while both being incomplete. Without sufficient aligned fraction, a high identity number may describe only a small common region and provide misleading species-level certainty.
When a report shows ANI, inspect the alignment fraction or coverage of the comparison and the exact reference genome. A classification boundary is part of a taxonomic method, not proof that two environmental populations are ecologically identical. They may differ in accessory genes or gene regulation. Keep identity, relatedness, strain variation and ecological behaviour conceptually separate. A single percentage is not a substitute for the full comparison context.
Dereplication removes redundant representatives by defined rules
A large soil study can reconstruct several similar MAGs from different samples or assembly runs. Dereplication groups genomes judged sufficiently similar under a chosen criterion and retains selected representatives. This makes catalogues easier to analyse, but it also hides some variation among related populations. The number of representatives after dereplication is not necessarily the number of species present in the soil, because recovery and similarity thresholds affect the count.
Check whether representatives were selected using completeness, contamination, size or other criteria. When comparing studies, find out whether they used similar dereplication thresholds and sample coverage. A catalogue reporting 100 representative MAGs may reflect a very different recovery pipeline from one reporting 200. Such counts are properties of recovered and filtered assemblies as much as properties of nature.
A MAG catalogue is not a census of soil species
Genome reconstruction favours organisms represented by enough high-quality DNA and distinguishable from similar populations. Rare taxa, high microdiversity, repeats and low coverage can prevent successful assembly. Many sequences may remain unbinned, and some legitimate organisms may not yield a usable MAG. Therefore the number of recovered genomes or estimated species should not be presented as the total richness of the underlying soil community.
An amplicon-based ASV survey and a MAG catalogue can give different pictures because they ask different questions and have different detection biases. The soil ASV guide explains read-based marker diversity. Use MAGs to build hypotheses about genomic capacities and associations, while recognising that a failure to assemble a genome does not prove that an organism was absent.
Functional annotations inherit genome-bin uncertainty
A reconstructed genome may contain genes suggesting particular metabolic potential. However, a contaminated bin can falsely assemble a pathway from contigs belonging to different organisms. An incomplete bin can omit real enzymes and make a capable organism appear deficient. A functional claim should therefore state the quality of the contributing bin and whether key genes reside on coherent, well-supported contigs. Strong biological stories need genomic context, not merely a long list of annotation names.
Suppose a draft MAG includes four of five expected genes in a proposed pathway. The fifth may genuinely be absent, or it may lie in a region not assembled or not assigned to the bin. Concluding that the organism cannot perform the pathway is premature. Equally, finding all five annotated genes does not establish expression or activity. Soil chemistry and physiological measurements remain necessary for process-rate claims.
Read mapping back to MAGs estimates representation with caveats
Researchers sometimes map reads from many soil samples to a reference MAG catalogue to estimate how strongly each reconstructed genome is represented. This can reveal patterns across treatments or depths. Yet coverage depends on the catalogue, sequence similarity, mapping thresholds and whether reads from close relatives align to the same reference. A representative MAG can absorb reads from multiple related strains, and a poor-quality reference may distort abundance estimates.
Inspect breadth of coverage, not only mean depth. A genome with reads covering most positions offers different evidence from one with high coverage restricted to a conserved gene. If several samples map to a MAG with very different breadths, a straightforward abundance comparison may be misleading. A sound report describes reference quality, unique-read requirements and normalisation, then restricts its ecological language to what those choices support.
Different denominators produce different abundance rankings
A study may display reads per kilobase of genome, coverage relative to total mapped reads, genome equivalents or a model-based relative abundance. The denominators differ. A larger recovered genome may collect more raw reads than a smaller genome at similar per-base coverage. A sample with substantial unbinned DNA may change the apparent fraction represented by each MAG. No number becomes absolute cells per gram merely because a heat map displays several decimal places.
For example, if one bin has 50,000 mapped bases across a 500,000-base reference and another has 100,000 mapped bases across a 2-million-base reference, their crude average depths are 0.1 and 0.05. The first has fewer mapped bases but twice the per-base depth. This classroom calculation ignores alignment and read-length complications; it demonstrates why genome length and denominator definitions must be considered.
Missing genes need cautious language
If a MAG is estimated at 70% completeness, a gene absent from that bin may simply fall in the missing 30% of expected genomic content. Even at 95% estimated completeness, uncertainty about the locations and functions of missing regions remains. A confident gene-absence claim may require a very high-quality assembly, robust annotation, targeted searches and comparative genomic context. The quality percentage alone cannot identify which genes are missing.
This matters for statements about auxotrophy, carbon pathways or nutrient transformations. A draft genome might appear unable to synthesise a compound because a key annotated gene was not recovered. Before proposing ecological dependence, check assembly gaps, alternative biochemical enzymes and functional annotation sensitivity. Phrase the result as ‘the gene was not detected in this draft bin under the stated methods’ when that is all the evidence supports.
A genome quality report needs a source-specific standard
Quality thresholds are useful screening rules, but they should not be applied without regard to organism biology and analysis goals. Bacterial and archaeal MAG standards are not automatically appropriate for every viral or eukaryotic assembly. Different biological systems require different markers and validation strategies. A result labelled 90% complete by one algorithm is not guaranteed to be numerically interchangeable with a result from another model.
The MIMAG reporting standard specifically addresses bacterial and archaeal genome reporting. If a dataset includes viruses, fungi or other eukaryotes, find the appropriate separate assessment framework. Do not repackage a general bacterial completeness percentage as universal genome completeness. Scope is one of the most important words in technical interpretation, particularly when charts aggregate unlike biological groups.
What to check when one MAG looks unusually successful
An extraordinary-quality bin deserves ordinary verification rather than automatic celebration. Inspect whether many contigs share consistent coverage, whether conserved marker genes support the reported lineage, and whether taxonomic assignments conflict. Review contamination metrics, strain variation, gene duplication and evidence for physical linkage of key genomic regions. If independent assembly methods or long reads support the result, note that stronger evidence specifically.
A very long contig could be a genuine near-complete chromosome or a misjoined sequence. A 99% completeness estimate can coexist with unresolved contamination or structural error. The question is not whether a quality score is impressive but whether different independent observations agree on the same reconstruction. Strong genome science often emerges from careful reconciliation of apparently inconsistent clues rather than from maximising a single metric.
Worked example: interpreting a questionable quality claim
A hypothetical report calls bin D a ‘complete, uncontaminated species genome’ and lists 92% estimated completeness, 4% contamination, 146 contigs and no RNA-gene evidence. Several aspects of that description are unjustified. The bin is not demonstrated complete; contamination is nonzero by the algorithm’s estimate; and the absence of RNA-gene assessment prevents an unqualified full high-quality MIMAG designation. Its species assignment still depends on the taxonomic method.
A cautious rewrite says: ‘Bin D is a fragmented draft MAG with an estimated 92% completeness and 4% contamination under the named quality tool. Additional required markers and assembly checks were not reported. Genome taxonomy was inferred using the stated reference.’ The revised text preserves what was measured and avoids claims that would need manual finishing or independent validation. This is technical translation serving accuracy rather than excitement.
The scientific consequences of false merges and false splits
A false merge combines contigs from distinct genomes and can invent gene combinations that no individual organism possesses. A false split divides one genome into multiple bins and can make a single organism appear as several incomplete populations. Both errors change abundance estimates and metabolic reconstructions. They can also affect later genome dereplication and phylogenetic placement, multiplying their influence across a catalogue.
If a reconstructed bin appears to possess an unusually broad set of functions or strangely conflicting taxonomy, a false merge becomes one possible explanation. If several small bins share nearly identical coverage and complementary markers, a false split may be worth investigating. Neither pattern is conclusive. Read the assembly graph and supporting evidence as available. The important principle is that computational grouping is a hypothesis about genomic ownership, not a guaranteed natural boundary.
Replication and ecology remain separate from genome quality
Even a well-assembled MAG from one soil sample does not establish that the represented population is consistently common across a landscape. Genome quality evaluates the reconstruction, not field generalisability. To compare treatment prevalence, readers need independent sample coverage, mapping consistency, ecological metadata and suitable statistical analysis. A study with one excellent bin can reveal new genomic possibilities while supporting only limited claims about prevalence.
Likewise, finding a gene in a MAG does not prove the organism expressed it during sampling. Environmental oxygen, substrate, temperature and regulation still control realised activity. Compare genome-resolved inferences with soil respiration and soil enzyme assays when the research question concerns processes rather than genomic potential.
Translate taxonomy and quality terms for general readers
Words such as ‘complete,’ ‘contaminated,’ ‘genome,’ ‘species’ and ‘coverage’ are familiar outside genomics but have specialised meanings inside a MAG report. A translated sentence can mislead if it changes ‘estimated completeness’ into ‘full genome’ or ‘contamination score’ into ‘laboratory sample was dirty.’ Include the method and uncertainty when using these terms. Translate the relationship, not merely each vocabulary item.
The cross-linguistic vocabulary guide and the context-and-meaning guide provide useful language-learning routes. In technical communication, define a term at first use and pair the explanation with its practical implication. Readers can understand a difficult genomic chart if the logical steps are made visible and the claims remain modest.
Frequently asked questions about soil MAGs
Is a MAG a cultivated bacterial isolate? No; its sequences are computationally recovered from mixed environmental DNA. Is completeness the exact percentage of a real chromosome reconstructed? No; it is an estimate under a particular assessment model. Does low contamination prove every contig has one origin? No; the quality tool can miss certain misassignments. Does high coverage guarantee correct assembly? No; repeats and closely related strains can produce errors even with many reads.
Can a medium-quality MAG still be useful? Yes, for analyses matched to its limitations. Does taxonomy remain fixed forever? No, classification can change with reference releases and improved evidence. If a pathway gene is missing, is the organism unable to perform the reaction? Not necessarily, particularly when assembly and annotation are incomplete. Can one MAG prove a soil treatment changed population abundance? No; ecological comparison needs sample-level replication and an appropriate measurement model.
Independent practice with explained answers
A hypothetical bin has estimated completeness 85%, contamination 3%, 90 contigs and mean coverage 12×. Can it be called a fully reconstructed genome? No: all four statistics describe a draft reconstruction and leave unresolved regions. Another bin has 94% completeness and 2% contamination but lacks the required RNA markers for a high-quality MIMAG classification. Should the report automatically label it a high-quality MIMAG? No; it should identify that only the numerical criteria are met while the additional requirements remain unverified.
What if two similar bins from different samples have high average nucleotide identity but low aligned fraction? Their similarity may be restricted to shared regions; more genomic evidence is needed before a confident species or strain conclusion. If a bin’s coverage rises after treatment, must cell numbers per gram have risen? No; extraction, mapping, sequencing and reference effects all require consideration. Each exercise illustrates why the denominator, biological unit and method belong in the final sentence.
Authoritative methods and further learning
For reporting criteria, consult MIMAG and MISAG standards. For a widely used genome-based taxonomy framework, read GTDB’s official description. For genome recovery and public-resource context, see MGnify’s environmental-DNA overview and NIH’s CheckM2 overview. The exact tool versions and study-specific parameter settings remain essential for interpreting a particular quality table.
For the wider learning route, combine this guide with shotgun functional profiling, soil qPCR copy-number interpretation and soil PLFA biomarkers. The three methods approach genomic potential, DNA-target quantity and biomass-related indicators from different directions rather than producing interchangeable counts.
The central lesson: a MAG is a testable genome hypothesis
A defensible MAG interpretation follows the evidence from soil to DNA extraction, reads, assembled contigs, proposed bins, quality markers, coverage patterns, taxonomy and biological inference. Completeness estimates how much of an expected genome may be recovered; contamination estimates the chance of mixed origin under a model; fragmentation and coverage describe other aspects of support. A beautiful report needs all these measures, not just a high percentage in its first column.
When communicating the result, say what the bin supports and what remains uncertain. A high-quality genome reconstruction can generate important hypotheses about previously uncultivated microorganisms, but it cannot turn genomic potential into measured ecological activity or one sampled population into a universal description of soil. That clear boundary makes the science both more rigorous and easier to teach.