Genome-Resolved Metagenomics: From Reads to MAGs in 10 Steps

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Key Takeaways

• Genome-resolved metagenomics reconstructs microbial genomes from mixed community sequencing data, bypassing traditional isolation methods, to produce Metagenome-Assembled Genomes (MAGs). This workflow systematically progresses from raw sequencing reads through quality control, assembly, binning, refinement, and quality assessment to generate MAGs for downstream ecological, functional, or evolutionary analyses.

• Critical quality control steps involve assessing raw read quality using tools like FastQC to identify adapter sequences and low-quality bases, followed by trimming to remove these artifacts, ensuring that at least 70-80% of reads are retained for downstream assembly. Metagenomic assembly strategies, such as using MEGAHIT for short reads or Flye for long reads, are chosen based on read length and community complexity, aiming for high N50 values and estimated completeness via conserved marker genes (e.g., using CheckM or BUSCO).

• Binning algorithms like MetaBAT2 or MaxBin2 leverage sequence composition (GC content, k-mer frequencies) and cross-sample coverage profiles to group contigs into putative genomes, with refinement steps employing taxonomic consistency checks (e.g., Kraken2) and coverage anomaly detection to merge or split bins. MAG quality is rigorously assessed using MIMAG standards, classifying them as draft (≥50% complete, <10% contamination) or high-quality draft (≥90% complete, <5% contamination).

• Taxonomic and functional annotation of MAGs, using tools like GTDB-Tk for taxonomy and Prokka for gene prediction, enables metabolic reconstruction and hypothesis generation regarding microbial community dynamics. For instance, functional annotation can reveal predicted metabolic interactions, such as the exchange of key amino acids between species, as demonstrated in infant oral microbiome studies.

• Data organization and submission to public repositories like NCBI's Sequence Read Archive or ENA are crucial for reproducibility, requiring meticulous documentation of sample metadata, sequencing parameters, and all intermediate and final analysis files. Tools like subMG are designed to streamline this submission process, ensuring FAIR data principles are met.

Metagenome-assembled genomes (MAGs) are reconstructed microbial genomes obtained directly from mixed community sequencing data instead of from isolated pure cultures. This workflow provides a ten-step protocol for researchers who have raw sequencing reads and need to produce quality-assessed MAGs for downstream ecological, functional, or evolutionary analysis. The workflow covers quality control, assembly, binning, refinement, quality assessment, taxonomic assignment, and data submission, with practical command examples and decision criteria at each stage.

At a Glance

The table below summarizes the ten steps, the primary tool category used at each stage, the key output produced, and the main quality decision a researcher must make before proceeding.

Step 1: Raw Data Acquisition and Organization

The starting point for any genome-resolved metagenomics project is a well-organized collection of raw sequencing reads. Most researchers receive FASTQ files from a sequencing facility or download them from public repositories. The National Center for Biotechnology Information (NCBI) maintains primary sequence databases, search systems, and analysis services that support the discovery and retrieval of metagenomic datasets for comparative work. Familiarity with these data resources is essential because public datasets often serve as validation sets or comparative references for newly generated MAGs.

Before beginning any computational work, create a directory structure that separates raw reads, intermediate files, and final outputs. A typical layout includes separate folders for raw data, trimmed reads, assemblies, bins, refined MAGs, and final quality reports. This organization prevents accidental overwriting and makes it easier to trace which parameters produced which results.

Record the following information for every sample in a plain-text metadata file: sample identifier, sequencing platform, read length, insert size, sequencing depth, collection site, collection date, and any relevant environmental parameters. This metadata becomes essential when you submit data to public repositories and when you interpret biological patterns across samples. The subMG tool was developed specifically to address the burden of assembling this metadata for metagenomics submissions, allowing researchers to input files and metadata in a single form and automating downstream tasks that otherwise require extensive manual effort and expertise.

Step 2: Read Quality Control and Trimming

Raw sequencing reads contain adapter sequences, low-quality bases, and potential contamination that can interfere with assembly and binning. Quality control removes these artifacts before assembly begins.

Start by running a quality assessment tool such as FastQC on all raw read files. Examine the per-base quality scores, GC content distribution, adapter content, and duplication levels. For paired-end reads, verify that read pairs are properly oriented and that the insert size distribution matches the library preparation protocol.

Trimming decisions depend on the observed quality profile. Common approaches include removing low-quality bases from read ends, clipping adapter sequences, and discarding reads that fall below a minimum length threshold after trimming. The specific thresholds you choose should reflect your sequencing platform and library preparation method. For example, Illumina data typically requires adapter clipping, while Oxford Nanopore or Pacific Biosciences long reads may require different quality filtering approaches.

After trimming, rerun the quality assessment to confirm that the filtering improved the data. Document the proportion of reads retained after trimming, as this metric helps identify problematic samples early. Samples with very low retention rates may indicate degraded DNA, sequencing failure, or contamination, and these samples may need to be re-sequenced or excluded from downstream analysis.

The Galaxy Training Network provides accessible workflow training and analysis tutorials that cover quality control procedures for metagenomic data. These tutorials are useful for researchers who want to see worked examples of quality assessment and trimming before applying the steps to their own data.

Step 3: Metagenomic Assembly

Assembly is the process of reconstructing longer contiguous sequences, called contigs, from overlapping sequencing reads. Metagenomic assembly is more challenging than single-genome assembly because the input contains multiple genomes at varying abundances, including closely related strains that share large regions of sequence identity.

The choice of assembler depends on read length and community complexity. Short-read assemblers such as MEGAHIT and metaSPAdes are designed for Illumina data and handle the uneven coverage typical of metagenomes. Long-read assemblers such as Flye and Canu can incorporate Oxford Nanopore or Pacific Biosciences reads, which often improve assembly continuity by spanning repetitive regions that short reads cannot resolve.

For hybrid approaches, researchers combine short and long reads to leverage the accuracy of short reads with the contiguity of long reads. This strategy is particularly valuable for complex communities where short-read assemblies remain fragmented. However, hybrid assembly requires more computational resources and careful parameter tuning.

Assembly parameters that require attention include k-mer sizes for short-read assemblers, minimum contig length thresholds for reporting, and coverage cutoffs for removing low-abundance contigs. The optimal parameters vary by dataset, so it is common practice to run multiple assemblies with different parameter sets and compare the results using assembly quality metrics.

The European Bioinformatics Institute (EMBL-EBI) provides training materials on bioinformatics data resources and practical analysis education, including content on assembly strategies for metagenomic data. These training resources help researchers understand the tradeoffs between different assembly approaches before committing computational time to a particular strategy.

Step 4: Assembly Quality Assessment

Assembly quality assessment determines whether the assembled contigs are suitable for binning. The two primary metrics are contiguity, measured by N50 and related statistics, and completeness, estimated by the presence of conserved single-copy marker genes.

N50 is the contig length at which half of the total assembly length is contained in contigs of that length or longer. Higher N50 values indicate more contiguous assemblies. However, N50 alone does not capture assembly correctness, and a highly contiguous assembly can still contain misjoins that create chimeric contigs.

Completeness estimation uses sets of genes that are expected to be present in single copy in most bacterial and archaeal genomes. The presence of these markers in an assembly indicates that the assembly captured most of the genome content. The absence of expected markers may indicate incomplete assembly or the presence of organisms with unusual gene content.

CheckM and BUSCO are commonly used tools for estimating completeness and contamination in metagenomic assemblies. Contamination is estimated by the presence of marker genes in more than one copy, which suggests that sequences from multiple organisms were assembled together.

Record the assembly statistics for each sample, including total assembled length, number of contigs, N50, and completeness estimates. These records allow you to compare assemblies across samples and to identify samples where assembly quality is insufficient for reliable binning. If completeness is low or contamination is high, consider adjusting assembly parameters, adding more sequencing data, or using a different assembler.

Step 5: Coverage Profiling

Coverage profiling calculates the depth of sequencing coverage for each contig in the assembly. Coverage values are used by binning algorithms to separate contigs from different populations, because contigs originating from the same genome tend to have similar coverage across samples.

To calculate coverage, map the quality-filtered reads back to the assembled contigs using a read aligner such as Bowtie2 or minimap2. The resulting alignment files are used to compute per-contig coverage values, typically expressed as average depth or reads per kilobase per million mapped reads (RPKM).

For multi-sample projects, coverage profiles across all samples provide a powerful signal for binning. Contigs from the same genome should show correlated coverage patterns across samples, reflecting the abundance of that genome in each sample. This differential coverage approach is particularly effective for separating genomes that are closely related but differ in abundance across environments.

Coverage values are also used to identify contigs that may be assembly artifacts. Contigs with extremely high coverage relative to the rest of the assembly may represent repetitive elements, circular elements, or contamination. Contigs with very low coverage may represent rare community members or sequencing errors that assembled by chance.

Store coverage values in a format that binning tools can read directly, such as a coverage table or a BAM file with depth information. Document the mapping parameters used, including the alignment tool version and any filtering thresholds applied to the alignments.

Step 6: Binning

Binning groups contigs into bins that represent putative genomes from individual populations. Binning algorithms use two main signals: sequence composition, such as GC content and k-mer frequencies, and coverage across samples.

Several binning tools are available, including MetaBAT2, MaxBin2, and CONCOCT. Each tool uses a different algorithm and may perform differently depending on the community composition and coverage profile. Running multiple binning tools and combining their results often produces better bins than relying on a single tool.

MetaBAT2 uses tetranucleotide frequency and coverage information to cluster contigs. MaxBin2 uses an expectation-maximization approach with marker genes to estimate genome completeness. CONCOCT uses Gaussian mixture models on composition and coverage features. The choice of tool should be guided by the characteristics of your dataset and the recommendations in the tool documentation.

After running the binning tools, examine the number and size distribution of the resulting bins. A typical metagenome may yield dozens to hundreds of bins, depending on community diversity and sequencing depth. Bins that are very small, containing only a few contigs, are unlikely to represent complete genomes and may be excluded from downstream analysis.

The nf-core documentation describes community pipeline standards for reproducible bioinformatics workflows, including pipelines that incorporate binning steps. These pipelines provide pre-configured workflows that handle the integration of multiple binning tools and produce standardized outputs.

Step 7: Bin Refinement

Bin refinement improves the quality of initial bins by merging bins that belong to the same genome, splitting bins that contain sequences from multiple genomes, and removing contaminating contigs.

The first refinement step is to assess the taxonomic consistency of each bin. Assign taxonomy to the contigs within each bin using a tool such as Kraken2 or CAT. If a bin contains contigs with conflicting taxonomic assignments, it may be contaminated and require splitting.

The second refinement step uses coverage information to identify contigs that do not follow the coverage pattern of the majority of the bin. Contigs with anomalous coverage may belong to a different population and should be removed.

The third refinement step uses marker gene analysis to identify bins with multiple copies of single-copy marker genes, which indicates contamination. Bins with high contamination can sometimes be split by separating contigs based on their coverage or composition signals.

Tools such as DAS Tool and MetaWRAP provide automated bin refinement workflows that integrate results from multiple binning tools and apply curation steps. These tools can merge bins from different binning algorithms that show consistent signals and split bins that show conflicting signals.

Document all refinement decisions, including which bins were merged, which were split, and which contigs were removed. This documentation is important for reproducibility and for interpreting the final MAG set.

Step 8: MAG Quality Assessment

MAG quality assessment uses completeness and contamination estimates to classify MAGs into quality categories. The minimum information about a metagenome-assembled genome (MIMAG) standards define three categories: draft MAGs have completeness greater than 50 percent and contamination less than 10 percent, high-quality draft MAGs have completeness greater than 90 percent and contamination less than 5 percent, and finished MAGs require additional manual curation to close gaps.

Completeness and contamination are estimated using single-copy marker genes. CheckM is the most widely used tool for this purpose, and it provides lineage-specific marker sets that improve the accuracy of estimates for different taxonomic groups. BUSCO is an alternative tool that uses a different set of marker genes and can be used for cross-validation.

The quality estimates have direct consequences for downstream analysis. MAGs with low completeness may produce biased functional profiles because missing genes are not detected. MAGs with high contamination may produce inflated estimates of genome size and may mix functional capabilities from different organisms.

Record the completeness and contamination estimates for every MAG in a summary table. This table should also include the number of contigs, total genome size, N50, and GC content for each MAG. These records allow you to filter MAGs based on quality thresholds appropriate for your research question.

For ecological studies that examine abundance patterns, only high-quality MAGs should be used for quantitative comparisons. For functional studies that examine gene content, MAGs with moderate completeness may still provide useful information, but the limitations should be acknowledged in the interpretation.

Step 9: Taxonomic and Functional Annotation

Taxonomic assignment places each MAG within the known tree of life, while functional annotation identifies the genes and metabolic pathways present in each MAG.

Taxonomic assignment for MAGs is typically performed using a set of conserved marker genes. Tools such as GTDB-Tk classify MAGs using the Genome Taxonomy Database, which provides a standardized taxonomy based on genome sequences. The resulting classifications can be compared with taxonomic assignments based on 16S rRNA gene sequences or other marker genes.

Functional annotation predicts the protein-coding genes in each MAG and assigns functional categories to those genes. Tools such as Prokka provide rapid annotation of bacterial and archaeal genomes, while more specialized tools can predict specific features such as carbohydrate-active enzymes, antibiotic resistance genes, or virulence factors.

The functional annotation of MAGs enables metabolic reconstruction and the prediction of ecological interactions. For example, a study of the infant oral microbiome used MAGs and genome-scale metabolic models to reveal genomic and functional characteristics of previously undescribed Streptococcus and Rothia species, predicting metabolic interactions including the exchange of key amino acids between these species. This type of analysis demonstrates how MAG-based functional predictions can generate testable hypotheses about microbial community dynamics.

The MGnify resource at EMBL-EBI provides tools and analysis approaches for genome-resolved metagenomics, including functional annotation of assembled metagenomes. The training materials from MGnify cover the processes used to annotate metagenomic data and are available for researchers who want to learn the standard approaches.

Step 10: Data Organization and Submission

Public data sharing is essential for reproducibility and for enabling large-scale comparative studies. Metagenomics datasets are often published with incomplete data because submission is cumbersome and time-consuming, requiring sample information, sequencing reads, assemblies, binned contigs, MAGs, and appropriate metadata.

The subMG tool simplifies and automates the submission of metagenomics study results to the European Nucleotide Archive (ENA). Researchers input files and metadata from their studies in a single form, and the tool automates downstream tasks that otherwise require extensive manual effort and expertise. The tool provides comprehensive documentation and example data for different use cases and can be operated via the command line or a graphical user interface.

Before submission, organize all files according to the repository requirements. This typically includes raw sequencing reads, quality-filtered reads, assemblies, bins, refined MAGs, and the metadata file. Verify that all sample identifiers are consistent across files and that the metadata includes all required fields.

The NCBI maintains multiple databases that accept metagenomic data, including the Sequence Read Archive for raw reads and GenBank for assembled sequences. The choice of repository depends on your funding requirements, institutional policies, and the preferences of your research community.

After submission, record the accession numbers for all submitted data. These accession numbers are cited in publications and allow other researchers to access your data for meta-analyses and comparative studies. Increased availability of well-documented and FAIR data benefits future research, particularly in meta-analyses and comparative studies.

Practical Implementation Steps

The following steps provide a practical sequence for implementing the ten-step workflow in a research project.

First, install the required software tools. Many tools are available through package managers such as Conda, which simplifies dependency management. The Bioconductor project provides official package and workflow documentation for reproducible genomic analysis in R, which may be useful for downstream statistical analysis of MAG data.

Second, create a project directory structure and metadata file before processing any data. This organization prevents errors and ensures that all outputs can be traced to their inputs.

Third, process one sample through the entire workflow before scaling to the full dataset. This pilot run identifies parameter issues and computational bottlenecks early, saving time and resources.

Fourth, document all software versions and parameters in a plain-text file. This documentation is essential for reproducibility and for troubleshooting when results are unexpected.

Fifth, run the workflow on the full dataset and monitor the outputs at each stage. Compare quality metrics across samples to identify outliers that may require additional attention.

Sixth, compile the final MAG quality table and filter MAGs based on the quality thresholds appropriate for your research question.

Seventh, submit all data to a public repository and record the accession numbers.

Records and Measurements

Maintain the following records throughout the genome-resolved metagenomics workflow.

The sample metadata file should contain all information needed to interpret the sequencing data, including sample identifiers, collection information, and sequencing parameters. This file is submitted to public repositories and is essential for data reuse.

The quality control log should record the number of raw reads, the number of reads retained after trimming, and the trimming parameters used for each sample. These records help identify samples with unusual quality profiles.

The assembly statistics table should record the total assembled length, number of contigs, N50, and completeness estimates for each assembly. These statistics allow comparison across samples and across assembly parameter sets.

The binning summary should record the number of bins produced by each binning tool, the number of bins retained after refinement, and the quality estimates for each final MAG.

The final MAG table should include the MAG identifier, completeness, contamination, genome size, number of contigs, N50, GC content, and taxonomic assignment. This table is the primary output of the workflow and is used for all downstream analyses.

Common Failure Patterns

Several recurring problems affect genome-resolved metagenomics projects. Recognizing these patterns early can save substantial time and computational resources.

Low sequencing depth produces fragmented assemblies and incomplete MAGs. If the assembly statistics show low N50 values and the completeness estimates are below 50 percent, additional sequencing may be required. Increasing sequencing depth is often necessary for complex communities with high diversity.

High contamination in bins indicates that sequences from multiple organisms were grouped together. This problem often arises when closely related strains are present in the community, because their shared sequence regions make it difficult to separate their genomes. Refinement steps that split bins based on coverage differences can help, but some communities may require strain-resolved approaches that are beyond the scope of standard MAG workflows.

Chimeric contigs, which contain sequences from different organisms, can arise during assembly when repetitive regions cause incorrect joins. These chimeras produce bins with mixed taxonomic signals and inflated contamination estimates. Removing contigs with conflicting taxonomic assignments during refinement can reduce this problem.

Parameter mismatches between the sequencing platform and the assembly tool can produce poor assemblies. For example, using a short-read assembler with long-read data, or vice versa, will produce suboptimal results. Verify that the assembly tool is appropriate for your read type before running the assembly.

Metadata errors, such as mismatched sample identifiers or missing collection information, can prevent data submission and complicate downstream analysis. Validate all metadata before beginning the workflow and again before submission.

Limitations and Interpretation Constraints

Genome-resolved metagenomics has inherent limitations that affect the interpretation of results.

MAGs are not complete genomes. Even high-quality draft MAGs may lack genes that are present in the actual organism, and the absence of a gene in a MAG does not prove that the organism lacks that gene. Functional conclusions based on MAG gene content should be treated as predictions that require experimental validation.

MAGs represent consensus sequences from populations, not individual cells. If a population contains multiple closely related strains, the MAG may represent a mosaic of sequences from those strains. This strain heterogeneity can affect both taxonomic assignment and functional predictions.

The completeness and contamination estimates are based on marker gene sets that may not be appropriate for all lineages. Unusual organisms with divergent gene content may produce inaccurate quality estimates. Cross-validation with multiple marker gene sets can help identify these cases.

Coverage-based abundance estimates derived from MAGs are relative, not absolute. The proportion of reads mapping to a MAG reflects the abundance of that population relative to the total community, but it does not provide absolute cell counts. Comparisons across samples should account for differences in sequencing depth and total community composition.

The computational requirements for genome-resolved metagenomics are substantial. Large datasets may require high-memory computing resources, and the workflow can take days or weeks to complete depending on the dataset size and available hardware. The Carpentries lessons provide foundational computing and data skills that help researchers manage these computational demands effectively.

Safety and Regulatory Context

Genome-resolved metagenomics does not involve the manipulation of hazardous biological materials, but researchers should be aware of relevant data handling and regulatory considerations.

When working with human-associated microbiome samples, researchers must comply with ethical and privacy requirements for human subjects data. This includes obtaining appropriate consent for data sharing and de-identifying sample information before submission to public repositories.

When working with pathogens or antimicrobial resistance genes, researchers should consider the potential dual-use implications of their findings. Some journals and funding agencies require review of research that could be misused to cause harm.

Data submission to public repositories must comply with the access policies of the repository. Some datasets may require controlled access due to privacy concerns, while others are freely available. The NCBI provides guidance on data submission policies and access controls for different data types.

Researchers should also be aware of the intellectual property considerations for data generated from environmental samples. Some countries have regulations governing the collection and use of genetic resources from their territories, and researchers should ensure compliance with applicable laws.

Professional Escalation Criteria

Researchers should seek additional expertise or escalate issues in the following situations.

If assembly quality remains poor after multiple parameter adjustments, consult with a bioinformatics specialist or the tool developers. Poor assembly quality may indicate fundamental issues with the sequencing data that require re-sequencing.

If MAG quality estimates are inconsistent across different assessment tools, seek guidance on the appropriate interpretation. Discrepancies between CheckM and BUSCO estimates may indicate unusual genome biology or issues with the marker gene sets.

If taxonomic assignments are uncertain or conflicting, consult with a taxonomy specialist. Some lineages are poorly represented in reference databases, and accurate classification may require phylogenetic analysis with additional markers.

If the computational requirements exceed available resources, seek access to high-performance computing facilities or cloud computing services. Many institutions provide support for researchers who need to scale up their computational capacity.

If data submission to public repositories fails due to metadata or format issues, contact the repository help desk for assistance. The subMG tool documentation and the repository submission guides provide troubleshooting information for common submission problems.

Decision Framework for MAG Selection and Downstream Analysis

The ten-step workflow produces a set of MAGs with associated quality metrics, but researchers still face the practical problem of deciding which MAGs to use for which downstream analyses. A structured decision framework prevents two common errors: including low-quality MAGs that distort biological conclusions and excluding useful MAGs that could support exploratory analysis. This section provides a tiered classification system, a scoring method for conflicting quality signals, and a decision matrix that links MAG quality to specific analysis types.

Tiered Classification System

Assign each MAG to one of four tiers based on completeness and contamination estimates from CheckM or BUSCO. The tiers extend beyond the MIMAG categories to provide finer resolution for practical decision making.

Tier 1 MAGs have completeness above 90 percent and contamination below 5 percent. These MAGs meet the high-quality draft standard and are suitable for all downstream analyses, including quantitative abundance comparisons, metabolic reconstruction, and comparative genomics. Tier 1 MAGs should be the primary focus of any study that makes claims about genome content or ecological function.

Tier 2 MAGs have completeness between 70 and 90 percent and contamination below 10 percent. These MAGs are suitable for presence-absence analysis, taxonomic assignment, and functional screening where the goal is to identify which genes or pathways are present instead of to quantify their abundance. Tier 2 MAGs should not be used for genome size estimation or for conclusions that depend on the absence of specific genes, because missing genomic regions may contain undetected functional content.

Tier 3 MAGs have completeness between 50 and 70 percent and contamination below 10 percent. These MAGs are useful for taxonomic discovery and for generating hypotheses about community membership. They may be included in phylogenetic analyses where the marker genes used for tree construction are present and complete. Tier 3 MAGs should be excluded from functional analyses that require near-complete gene inventories.

Tier 4 MAGs have completeness below 50 percent or contamination above 10 percent. These MAGs fail the minimum MIMAG draft standard and should not be reported as MAGs in publications. They may still contain useful information for targeted analyses, such as the presence of a specific gene of interest, but they must be described as partial bins instead of MAGs.

Scoring Method for Conflicting Quality Signals

Completeness and contamination estimates can conflict with other quality indicators, and researchers need a systematic method to resolve these conflicts. Use a five-point scoring system that integrates multiple lines of evidence beyond the marker gene estimates.

Score one point for each of the following criteria that the MAG meets. First, the completeness estimate from the primary tool, CheckM or BUSCO, exceeds the tier threshold. Second, a second independent marker gene tool confirms the completeness estimate within 10 percentage points. Third, the GC content distribution across the MAG contigs is unimodal, meaning the contigs show a single dominant GC peak instead of multiple peaks that suggest mixed genomes. Fourth, the coverage profile of the MAG contigs is consistent across samples, with no contigs showing coverage values that deviate by more than fivefold from the median coverage of the MAG. Fifth, the taxonomic assignments of the individual contigs within the MAG are consistent at the phylum level, with no more than 5 percent of contigs assigned to a different phylum than the majority.

A MAG that scores five points has strong support across all quality dimensions and can be used with confidence. A MAG that scores three or four points has moderate support and should be used with caution, with the specific weakness documented in the analysis. A MAG that scores two points or fewer has conflicting quality signals and should be downgraded one tier from its marker gene based classification.

This scoring method is particularly valuable when marker gene estimates suggest high quality but other signals indicate problems. For example, a MAG with 95 percent completeness and 2 percent contamination but a bimodal GC distribution likely contains sequences from two organisms with different base compositions. The scoring method flags this inconsistency and prevents the MAG from being used in quantitative analyses despite its favorable marker gene estimates.

Decision Matrix for Analysis Types

The decision matrix below links MAG tiers to specific downstream analysis types. This matrix provides a practical reference for planning analyses before the workflow begins and for documenting analysis choices in publications.

For taxonomic discovery and community composition analysis, Tier 1, Tier 2, and Tier 3 MAGs are all acceptable. Taxonomic assignment relies on a small set of conserved marker genes that are often present even in incomplete MAGs. The inclusion of lower-tier MAGs increases the sensitivity of community composition analyses by capturing rare or poorly assembled populations.

For phylogenetic analysis, Tier 1 and Tier 2 MAGs are recommended. Phylogenetic trees are sensitive to missing data, and incomplete MAGs can produce long branches or unstable placements. If Tier 3 MAGs are included in phylogenetic analysis, the tree should be built using only the marker genes that are present in all included MAGs, and the results should be interpreted with caution.

For functional gene content analysis, Tier 1 MAGs are required when the analysis concludes that a gene is absent from a genome. Absence conclusions are only valid when the genome is nearly complete, because missing genes may reside in unassembled genomic regions. Tier 2 MAGs can be used for presence analysis, where the goal is to identify which genes are present, but absence conclusions from Tier 2 MAGs should be explicitly flagged as uncertain.

For metabolic reconstruction and genome-scale modeling, Tier 1 MAGs are required. Metabolic models depend on complete or near-complete gene inventories to predict metabolic capabilities and interactions. A study of the infant oral microbiome used MAGs and genome-scale metabolic models to reveal genomic and functional characteristics of previously undescribed Streptococcus and Rothia species, predicting metabolic interactions including the exchange of key amino acids between these species. This type of analysis demonstrates how MAG-based functional predictions can generate testable hypotheses about microbial community dynamics, but the reliability of these predictions depends on the quality of the underlying MAGs.

For quantitative abundance comparisons across samples, Tier 1 MAGs are required. Abundance estimates based on read mapping are distorted when the reference MAG is incomplete, because reads from missing genomic regions cannot map to the MAG. This distortion is systematic and can produce false differences in abundance between samples if the completeness of a MAG varies across samples due to differences in sequencing depth.

For comparative genomics across many MAGs, Tier 1 and Tier 2 MAGs are acceptable when the comparison focuses on gene presence and absence patterns. The pangenome analysis should account for the different completeness levels of the included MAGs, typically by clustering genes into orthologous groups and treating missing genes as unknown instead of absent.

Implementation Steps for the Decision Framework

Apply the decision framework in four steps after the MAG quality assessment step and before any downstream analysis.

First, compile the quality metrics for all MAGs into a single table that includes completeness, contamination, genome size, number of contigs, N50, GC content, and taxonomic assignment. This table is the input for the tier classification.

Second, assign each MAG to a tier using the completeness and contamination thresholds described above. Record the tier assignment in the MAG table.

Third, apply the five-point scoring method to all Tier 1 and Tier 2 MAGs. This step identifies MAGs with conflicting quality signals that should be downgraded despite favorable marker gene estimates. Record the score and any downgrades in the MAG table.

Fourth, use the decision matrix to select the MAG set for each planned downstream analysis. Document the tier threshold applied for each analysis in the methods section of any publication or report.

Records for the Decision Framework

Maintain a decision log that records the tier assignment, scoring results, and analysis selection for each MAG. This log should include the version of the marker gene tool used for completeness and contamination estimates, the parameters used for the scoring method, and the date of the assessment.

The decision log serves two purposes. First, it provides transparency for reviewers and readers who need to understand which MAGs were used for which analyses and why. Second, it allows the analysis to be repeated with different quality thresholds to test whether the biological conclusions are robust to MAG selection choices. Sensitivity analysis of this type strengthens the conclusions of any genome-resolved metagenomics study.

Common Failure Patterns in MAG Selection

Several recurring problems affect MAG selection decisions. Recognizing these patterns prevents the inclusion of inappropriate MAGs in downstream analyses.

The first pattern is using all MAGs for all analyses without applying quality thresholds. This approach produces inflated estimates of community diversity and distorts functional profiles by including partial genomes with incomplete gene inventories. The decision matrix prevents this error by linking analysis types to minimum quality tiers.

The second pattern is applying a single quality threshold to all analyses regardless of the analysis type. A threshold that is appropriate for quantitative abundance comparisons may be unnecessarily strict for taxonomic discovery, excluding useful information about rare community members. The tiered system allows different thresholds for different analysis types.

The third pattern is relying solely on marker gene completeness and contamination estimates without checking other quality signals. The five-point scoring method catches problems that marker gene estimates miss, such as chimeric assemblies or mixed genomes with unusual GC content distributions.

The fourth pattern is failing to document MAG selection decisions. Publications that do not state which MAGs were used for which analyses and why are difficult to evaluate and reproduce. The decision log provides the documentation needed for transparent reporting.

Limitations of the Decision Framework

The decision framework has limitations that should be acknowledged when applying it to research data.

The tier thresholds are based on the MIMAG standards, which were developed for bacterial and archaeal genomes. Eukaryotic MAGs, such as those from microeukaryotes in environmental samples, may require different thresholds because their genomes have different marker gene content and ploidy levels. Researchers working with eukaryotic MAGs should consult the relevant literature for appropriate quality standards.

The five-point scoring method uses GC content distribution and coverage consistency as quality indicators. These indicators are less informative for communities with many closely related strains, where the coverage profiles of different strains may be similar and the GC content distributions may overlap. In these cases, the scoring method may not detect contamination that is present.

The decision matrix provides general guidance, but specific research questions may require different thresholds. For example, a study focused on a single gene of interest may use a Tier 4 bin for targeted analysis, while a study making broad claims about community function should restrict analysis to Tier 1 MAGs. Researchers should adapt the framework to their specific research context while documenting any deviations from the standard thresholds.

The computational requirements for applying the scoring method are modest, but the method requires access to the per-contig coverage values and taxonomic assignments that are generated during the binning and refinement steps. Researchers who did not save these intermediate files during the initial workflow may need to rerun the coverage profiling and taxonomic assignment steps to apply the scoring method.

Frequently Asked Questions

What is the minimum sequencing depth required for MAG reconstruction?

The required sequencing depth depends on community complexity and the abundance of the target organisms. Communities with high diversity or many closely related strains require more depth to achieve complete MAGs. A practical approach is to assess completeness estimates from an initial assembly and increase sequencing if completeness is below 50 percent for the populations of interest.

How do I choose between short-read and long-read sequencing for metagenomics?

Short-read sequencing provides high accuracy and is well supported by established assembly and binning tools. Long-read sequencing produces more contiguous assemblies but has higher error rates and higher cost. Hybrid approaches that combine both read types can improve assembly quality for complex communities. The choice depends on your research question, budget, and access to sequencing platforms.

What is the difference between a MAG and an isolate genome?

An isolate genome is sequenced from a pure culture of a single organism, while a MAG is reconstructed from mixed community sequencing data. Isolate genomes are generally more complete and accurate because they do not contain sequences from other organisms. MAGs enable the study of organisms that cannot be cultured, but they have limitations in completeness and may represent population consensus sequences instead of individual genomes.

How do I know if my MAGs are good enough for publication?

The MIMAG standards provide minimum quality thresholds for reporting MAGs. Draft MAGs require completeness greater than 50 percent and contamination less than 10 percent. High-quality draft MAGs require completeness greater than 90 percent and contamination less than 5 percent. Journals may have additional requirements, so check the specific journal guidelines before submission.

Can I compare MAGs across different studies?

Comparisons across studies are possible when the MAGs are generated using comparable methods and quality thresholds. Differences in assembly, binning, and quality assessment methods can introduce biases that complicate comparisons. Public repositories that store MAGs with standardized metadata facilitate cross-study comparisons, and the increasing availability of well-documented data supports meta-analyses and comparative studies.

What should I do if my bins have high contamination?

High contamination indicates that sequences from multiple organisms were grouped together. Try refinement steps that split bins based on coverage differences or taxonomic signals. If contamination persists, consider using a different binning tool or combining results from multiple tools. In some cases, the community may contain closely related strains that cannot be separated with standard binning approaches.

How long does the genome-resolved metagenomics workflow take?

The time required depends on dataset size, community complexity, and available computational resources. A small dataset with a few samples may take a few days, while large datasets with many samples can take weeks. The assembly and binning steps are the most computationally intensive. Running a pilot sample through the full workflow provides a time estimate for the complete dataset.

What are the most common mistakes in MAG reconstruction?

The most common mistakes include insufficient sequencing depth, inappropriate assembly parameters, failure to document metadata, and inadequate quality assessment. These mistakes produce MAGs that are incomplete, contaminated, or poorly documented, which limits their utility for downstream analysis and data sharing. Following a structured workflow with quality checks at each stage reduces these errors.

Related Bioinformatics Guides

• Metagenomics Data Analysis: From Raw Reads to Biological Insights
• RNA-Seq Data Analysis Workflow: From Raw Reads to Insights
• RNA Sequencing Data Analysis: From Raw Reads to Differential Expression
• Metagenome Assembled Genome Analysis: From Bins to Biological Insights
• Metagenomics Pipeline: From Raw Reads to Taxonomic and Functional Profiles

Related Clinical & Scientific Guides

• A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data
• Computational Immunology: Modeling the Immune System
• How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices

References and Further Reading

• NCBI Data Resources. National Center for Biotechnology Information.
• EMBL-EBI Training. European Bioinformatics Institute.
• Bioconductor. Bioconductor Project.
• Galaxy Training Network. Galaxy Project.
• nf-core Documentation. nf-core.
• The Carpentries Lessons. The Carpentries.
• subMG automates data submission for metagenomics studies.. 2025.
• Metagenomics bioinformatics at MGnify. 2023.
• Unique ecology of co-occurring functionally and phylogenetically undescribed species in the infant oral microbiome.. 2026.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.