# Understanding Assembly Metrics: What N50, L50, and BUSCO Really Tell You About Genome Quality Genome assembly quality metrics answer a practical question: can you trust the sequence you are about to analyze? N50, L50, and BUSCO scores are the most frequently reported numbers in assembly projects, yet they measure different aspects of quality and can mislead when interpreted alone. N50 and L50 describe contiguity, or how much of the genome is contained in the largest pieces. BUSCO describes completeness and redundancy relative to expected gene content. A highly contiguous assembly can still be missing genes, and a complete assembly can be fragmented into thousands of pieces. Researchers working with long-read sequencing data, including nanopore and PacBio workflows, need to read these metrics together to decide whether an assembly supports downstream analysis such as structural variant calling, gene annotation, or comparative genomics. This article explains what each metric calculates, what it cannot tell you, and how to combine them into a practical quality assessment for your own assemblies. ## The Problem with Single-Number Quality Scores Assembly quality is multidimensional. No single metric captures whether a genome is correctly assembled, complete, or free of misjoins. The most common mistake in genome projects is treating N50 as a proxy for overall quality. N50 reports the length at which half the assembly is contained in contigs or scaffolds of that size or larger. It says nothing about base-level accuracy, gene content, or whether the order and orientation of sequences match the true genome. Consider two assemblies of the same species. One has an N50 of 10 megabases but is missing 30 percent of expected genes. Another has an N50 of 1 megabase but contains 98 percent of expected genes in single copy. For most biological questions, the second assembly is more useful, even though its contiguity looks worse on paper. The reverse situation also occurs. An assembly can have excellent gene completeness but be so fragmented that structural variant analysis becomes impractical. The practical implication is that you need a small set of complementary metrics. Contiguity metrics such as N50 and L50 tell you about the physical structure of the assembly. Completeness metrics such as BUSCO tell you about the gene space. Together they give you a two-dimensional view of assembly quality. The [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible workflows that walk through assembly quality assessment, which is useful for researchers who want to see these metrics computed on real data before applying them to their own projects. ## What N50 Measures and What It Misses N50 is a weighted median statistic. To calculate it, you sort all contigs or scaffolds from longest to shortest. You then add lengths from the top until you reach 50 percent of the total assembly length. The length of the contig that brings you to that halfway point is the N50. If your assembly is 100 megabases and your N50 is 5 megabases, then half of the assembled sequence is contained in contigs of 5 megabases or longer. L50 is the number of contigs needed to reach that halfway point. If your N50 is 5 megabases and your L50 is 10, then 10 contigs account for half of the assembly. L50 gives you a sense of how many large pieces you are working with. A low L50 relative to the total number of contigs indicates that a few large pieces dominate the assembly. A high L50 indicates that the assembly is more evenly fragmented. These metrics are useful for comparing assemblies from the same species or from species with similar genome sizes. They are less useful for comparing across species with very different genome sizes. A bacterial genome with an N50 of 4 megabases may be a single complete chromosome. A plant genome with an N50 of 4 megabases may represent only a small fraction of the total sequence. The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) host thousands of assemblies with reported N50 values, and browsing these records shows the wide range of contiguity across taxa. N50 does not measure correctness. A contig can be chimeric, meaning it joins sequences from different parts of the genome, and still contribute to a high N50. Misassemblies at repeat regions are common in both short-read and long-read assemblies. The N50 statistic has no way to detect these errors. You need additional evidence such as read mapping consistency, optical maps, or Hi-C data to validate the physical structure of the assembly. Another limitation is that N50 is sensitive to how you define the assembly. Scaffolds have higher N50 values than contigs because scaffolding joins contigs with gaps. Reporting scaffold N50 without specifying that gaps are present can inflate the apparent quality. Always check whether the reported N50 refers to contigs or scaffolds before comparing values. ## BUSCO Completeness and Redundancy BUSCO, which stands for Benchmarking Universal Single-Copy Orthologs, takes a different approach. Instead of measuring physical contiguity, it measures how many genes from a conserved ortholog set are present in your assembly. The method relies on the observation that certain genes are present in single copy across a broad range of species. These genes are unlikely to be lost or duplicated in normal evolution, so their presence in an assembly indicates that the corresponding genomic regions were successfully assembled. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes how the tool uses datasets of near-universal single-copy orthologs derived from the OrthoDB database. When you run BUSCO on an assembly, it searches for these conserved genes and classifies each one as complete, duplicated, fragmented, or missing. The output gives you four percentages that sum to the total number of BUSCO groups searched. A complete BUSCO gene is present in the assembly and matches the expected length. A duplicated BUSCO gene is present in more than one copy. Duplication can indicate a true recent whole-genome duplication, but it can also indicate a haplotype that was not collapsed during assembly. A fragmented BUSCO gene is present but broken across multiple contigs. A missing BUSCO gene is absent from the assembly entirely. The choice of BUSCO lineage dataset matters. You must select the dataset that matches your organism. Using a vertebrate dataset on a fungal assembly will produce misleading results because many vertebrate-specific genes will be missing. The [BUSCO documentation](https://pubmed.ncbi.nlm.nih.gov/34936221) emphasizes that the auto-lineage workflow can detect the appropriate dataset automatically, which is useful when you are unsure of the taxonomic placement of your sample. BUSCO scores are biologically meaningful because they reflect the gene space instead of the physical structure. A fragmented assembly can still have high BUSCO completeness if the conserved genes happen to fall within the larger contigs. Conversely, a highly contiguous assembly can have low BUSCO completeness if the assembly is missing large genomic regions or if the gene prediction step failed. ## How N50, L50, and BUSCO Work Together The three metrics answer different questions. N50 and L50 answer the question of how much sequence is in the largest pieces. BUSCO answers the question of how much of the expected gene content is present and in the correct copy number. A complete quality assessment requires both types of information. A useful way to think about the combination is to consider four scenarios. First, high N50 with high BUSCO completeness indicates a strong assembly that is both contiguous and gene-complete. This is the target for most projects. Second, high N50 with low BUSCO completeness suggests that the assembly is physically large but missing genomic regions. This can happen when repetitive regions collapse or when some chromosomes are underrepresented in the sequencing data. Third, low N50 with high BUSCO completeness indicates a fragmented but gene-complete assembly. This assembly may be usable for gene-based analyses but difficult for structural variant calling. Fourth, low N50 with low BUSCO completeness indicates a poor assembly that likely needs more sequencing or a different assembly strategy. The [European mink genome project](https://pubmed.ncbi.nlm.nih.gov/37834264) provides an example of a high-quality assembly that combines both types of metrics. The assembly achieved a chromosome-level structure with an N50 of 154.1 megabases and a BUSCO completeness score of 98.2 percent. The combination of high contiguity and high completeness made this assembly suitable for conservation genomics applications such as management unit revision and captive breeding impact assessment. The [root-lesion nematode genome project](https://pubmed.ncbi.nlm.nih.gov/37108472) shows a different profile. The assembly had an N50 of 1.72 megabases and a BUSCO completeness of 65.4 percent, with 32.8 percent of BUSCOs missing. The authors noted that the assembly was generated from only 500 nematodes using an ultra-low DNA input workflow. The lower completeness reflects the difficulty of assembling a genome from limited starting material. This example illustrates that BUSCO completeness can flag problems that N50 alone would miss. ## At a Glance | Metric | What It Measures | What It Misses | Interpretation Guidance | |--------|-----------------|----------------|------------------------| | N50 | Contiguity: length at which half the assembly is in contigs or scaffolds of that size or larger | Base accuracy, gene content, misassembly errors, order and orientation | Report contig N50 and scaffold N50 separately. Compare only within similar genome sizes. | | L50 | Number of contigs or scaffolds needed to reach half the assembly length | Gene content, completeness, duplication status | Low L50 relative to total contig count indicates a few dominant pieces. High L50 indicates fragmentation. | | BUSCO completeness | Percentage of expected single-copy orthologs present in the assembly | Physical contiguity, base accuracy, non-coding regions | Use the correct lineage dataset. Report complete, duplicated, fragmented, and missing percentages. | ## Practical Workflow for Assembly Quality Assessment A reproducible quality assessment workflow should include the following steps. First, run a contiguity analysis using a tool such as QUAST, which reports N50, L50, and related statistics. Second, run BUSCO with the appropriate lineage dataset. Third, map the raw reads back to the assembly to check for coverage uniformity and to identify potential misassemblies. Fourth, if Hi-C data are available, check whether the assembly can be clustered into expected chromosome numbers. The [AquaaG pipeline](https://doi.org/10.1016/j.mex.2026.103955) integrates these steps into an automated workflow. It retrieves assemblies from NCBI, runs QUAST for assembly quality assessment, runs BUSCO for gene-space completeness evaluation, and then proceeds to annotation. The pipeline is configured through YAML files and supports species-level, kingdom-level, and custom assembly-based analyses. This type of automated workflow reduces the risk of inconsistent quality assessment across multiple assemblies. For researchers who prefer to run individual tools, the [Bioconductor project](https://bioconductor.org/) offers R packages for genomic analysis that can complement assembly quality assessment. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide structured learning pathways for bioinformatics analysis, including assembly and quality assessment modules. The [nf-core documentation](https://nf-co.re/docs) describes community pipelines that follow reproducibility standards, which is valuable when you need to share your analysis with collaborators or reviewers. The [Carpentries lessons](https://carpentries.org/lessons) cover foundational computing skills such as shell scripting and data management. These skills are necessary for running assembly quality tools efficiently, especially when working with large eukaryotic genomes that require substantial computational resources. ## Records and Measurements for Assembly Quality Keeping systematic records of assembly quality metrics is essential for reproducibility and for comparing assemblies across projects. At minimum, record the following for each assembly: total assembly length, number of contigs and scaffolds, contig N50, scaffold N50, L50, GC content, BUSCO lineage dataset used, and the four BUSCO percentages. Also record the software versions for the assembler and the quality assessment tools, since different versions can produce slightly different results. The [European mink assembly](https://pubmed.ncbi.nlm.nih.gov/37834264) reports a genome size of 2586.3 megabases across 25 scaffolds, with an N50 of 154.1 megabases and a GC content of 41.9 percent. The assembly was generated using PacBio HiFi long reads and Hi-C data for chromosome scaffolding. These records allow other researchers to understand the assembly structure and to compare it with other mammalian genomes. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) provides another example of detailed record keeping. The assembly contains two haplotypes with total lengths of 868.84 and 869.52 megabases. Most of haplotype 1 is scaffolded into 30 chromosomal pseudomolecules, including the Z sex chromosome. The mitochondrial genome was assembled separately at 15.38 kilobases. Reporting haplotype-level information is important for species with high heterozygosity, where haplotypes may not collapse cleanly during assembly. For your own projects, create a quality assessment table that includes all of these fields. Update the table whenever you generate a new assembly or when you re-run quality assessment with updated software. This record becomes part of the assembly metadata that you should submit to public databases such as NCBI when you deposit your assembly. ## Common Failure Patterns in Assembly Quality Several failure patterns recur across genome assembly projects. Recognizing these patterns helps you diagnose problems and decide whether to invest in more sequencing or a different assembly strategy. The first pattern is high N50 with low BUSCO completeness. This often indicates that the assembly is missing large genomic regions. The missing regions may be repetitive, GC-rich, or otherwise difficult to sequence. In the [root-lesion nematode assembly](https://pubmed.ncbi.nlm.nih.gov/37108472), the BUSCO completeness of 65.4 percent with 32.8 percent missing suggests that a substantial portion of the gene space was not assembled. The authors used an ultra-low DNA input workflow, which may have limited the sequencing depth and coverage uniformity. The second pattern is high BUSCO completeness with low N50. This indicates that the gene space is well represented but the assembly is fragmented. Fragmentation can result from insufficient sequencing depth, high repeat content, or an assembler that is conservative about joining contigs. This assembly may be usable for gene-based analyses but will be challenging for structural variant calling or chromosome-level comparisons. The third pattern is high BUSCO duplication. When BUSCO reports a high percentage of duplicated genes, it may indicate that haplotypes were not collapsed during assembly. This is common in heterozygous species or in assemblies that use trio binning or other haplotype-aware strategies. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) explicitly reports two haplotypes, which is appropriate for a heterozygous individual. If you are aiming for a haploid representation of the genome, high duplication indicates that you need to apply haplotype collapsing or purging steps. The fourth pattern is low completeness across all metrics. This indicates a fundamentally poor assembly that likely needs more sequencing data, a different library preparation, or a different assembler. Before discarding the assembly, check whether the BUSCO lineage dataset is appropriate for your organism. Using the wrong lineage will produce artificially low completeness. ## Limitations of Assembly Metrics Every assembly metric has limitations that affect interpretation. N50 and L50 are sensitive to the total assembly size and to the definition of contigs versus scaffolds. They do not measure base-level accuracy, and they cannot detect misjoins or chimeric contigs. A high N50 can coexist with serious structural errors. BUSCO completeness is limited to the gene space defined by the ortholog set. It does not assess non-coding regions, regulatory elements, or repetitive sequences. A genome can have high BUSCO completeness and still be missing large portions of the actual genome. BUSCO also depends on the quality of the gene prediction or the search parameters. Fragmented genes may be missed if the search sensitivity is too low. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) notes that BUSCO can assess different data types, including genome assemblies, transcriptomes, annotated gene sets, and metagenome-assembled genomes. The interpretation of BUSCO results depends on the data type. For a transcriptome, missing BUSCOs may reflect low expression instead of assembly problems. For a metagenome-assembled genome, missing BUSCOs may reflect the incompleteness of the bin instead of the quality of the assembly. Another limitation is that assembly metrics are not directly comparable across species with different genome biology. Species with high repeat content, such as many plant genomes, will have lower N50 values than species with low repeat content, even when both assemblies are of similar quality. Species that have undergone recent whole-genome duplication will have higher BUSCO duplication than species that have not. Always compare assemblies within a taxonomic context. ## Structural Variant Calling and Assembly Quality Structural variant calling from long-read data depends heavily on assembly quality. If you are using an assembly as a reference for structural variant detection, the contiguity and completeness of the reference directly affect your ability to detect deletions, insertions, inversions, and duplications. A fragmented reference with many small contigs makes it difficult to map reads across variant breakpoints. A reference with missing regions will cause variants in those regions to be missed entirely. The [European mink assembly](https://pubmed.ncbi.nlm.nih.gov/37834264) demonstrates the value of a chromosome-level reference for population genomics. With more than 99.9 percent of the assembly clustered into 20 chromosomal pseudomolecules, researchers can map reads to complete chromosomes and detect structural variants with confidence. The high BUSCO completeness of 98.2 percent indicates that the gene space is well represented, which is important for linking structural variants to gene function. For structural variant calling, you should also consider the read depth and coverage uniformity of the sequencing data used to generate the assembly. Low coverage regions will produce gaps in the assembly, and these gaps will be invisible to N50 and BUSCO metrics. Mapping the raw reads back to the assembly and checking coverage across the genome is a necessary complement to assembly quality metrics. ## Choosing Between Assembly Strategies Based on Quality Metrics The choice of sequencing platform and assembly strategy depends on your quality targets. PacBio HiFi reads produce highly accurate long reads that are well suited for generating contiguous assemblies. The [European mink project](https://pubmed.ncbi.nlm.nih.gov/37834264) used PacBio HiFi reads and achieved chromosome-level assembly with Hi-C scaffolding. The [root-lesion nematode project](https://pubmed.ncbi.nlm.nih.gov/37108472) also used PacBio HiFi reads but with an ultra-low DNA input workflow, which produced a more fragmented assembly. Nanopore sequencing offers longer reads than PacBio HiFi but with lower per-base accuracy. Nanopore assemblies may require additional polishing steps to achieve base-level accuracy. The choice between platforms involves tradeoffs between read length, accuracy, throughput, and cost. Your quality metrics will reflect these tradeoffs, and you should interpret them in the context of your sequencing strategy. For species with high heterozygosity, consider whether you need a haploid representation or a haplotype-resolved assembly. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) reports two haplotypes, which is appropriate for a heterozygous individual. Haplotype-resolved assemblies are more complex to generate but provide more complete information for population genomics and structural variant analysis. ## Professional Escalation Criteria Knowing when to escalate an assembly quality problem is important for avoiding wasted effort on downstream analyses. Escalate to additional sequencing or a revised assembly strategy when you observe any of the following conditions. First, BUSCO completeness is below 90 percent for a eukaryotic genome and the missing BUSCOs are distributed across the genome instead of concentrated in a few regions. Second, BUSCO duplication is above 10 percent and you intended to produce a haploid assembly. Third, the N50 is more than 10 times smaller than the expected chromosome size for your species and you need chromosome-level resolution for your analysis. Fourth, read mapping reveals large regions of zero coverage that are not explained by known repeat content. Before escalating, verify that your BUSCO lineage dataset is correct and that your quality assessment tools are up to date. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes the auto-lineage workflow that can help you select the appropriate dataset. Also verify that your assembly software and parameters are appropriate for your data type. Different assemblers perform differently on different genomes, and switching assemblers may resolve quality problems without additional sequencing. When you escalate, document the specific metrics that triggered the escalation and the steps you took to verify the problem. This documentation is valuable for troubleshooting and for reporting to collaborators or reviewers. The [nf-core documentation](https://nf-co.re/docs) emphasizes reproducibility standards that include detailed logging of parameters and versions, which supports this type of troubleshooting. ## Safety and Data Management Context Genome assembly projects involve large data volumes that require careful management. Raw sequencing data, intermediate files, and final assemblies should be stored with clear naming conventions and version control. The [Carpentries lessons](https://carpentries.org/lessons) cover data management practices that are directly applicable to genome projects. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide additional guidance on data organization and analysis workflows. When you deposit your assembly in a public database such as NCBI, include the quality metrics in the assembly metadata. The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) provide structured fields for assembly statistics, and including these metrics helps other researchers assess the quality of your assembly before using it. The [AquaaG pipeline](https://doi.org/10.1016/j.mex.2026.103955) demonstrates how assembly quality assessment can be integrated into a reproducible workflow that produces the metadata needed for database submission. For projects involving endangered species or sensitive samples, consider the ethical and legal context of your sequencing data. The [European mink project](https://pubmed.ncbi.nlm.nih.gov/37834264) highlights the conservation applications of genome sequencing, but it also raises questions about data sharing and sample provenance. Consult with your institution and any relevant regulatory bodies before depositing sensitive data. ## A Practical Decision Framework for Triaging Assembly Quality Problems Quality metrics become useful only when they drive a decision. Researchers often collect N50, L50, and BUSCO values without a structured way to act on them. A practical decision framework helps you determine whether your assembly is ready for downstream analysis, needs targeted improvement, or requires a fundamental restart. This section provides a stepwise triage system that combines contiguity and completeness metrics with read mapping evidence to produce clear action items. ### Step 1: Establish Your Quality Thresholds Before Assessment The first mistake in assembly quality assessment is deciding what is acceptable after you see the results. This creates confirmation bias and makes it difficult to compare assemblies objectively. Set your thresholds before running the assessment, based on your downstream analysis requirements and the biology of your species. For a eukaryotic genome intended as a reference for population genomics or structural variant calling, a reasonable starting threshold is 90 percent complete BUSCO genes with less than 5 percent duplication. The [European mink assembly](https://pubmed.ncbi.nlm.nih.gov/37834264) achieved 98.2 percent BUSCO completeness with chromosome-level scaffolding, which represents the quality target for conservation genomics applications. For a draft assembly intended for gene discovery or comparative genomics, 80 percent completeness may be acceptable, but you should document the limitation in any downstream interpretation. For prokaryotic genomes, the thresholds differ because the genomes are smaller and the expectation for completeness is higher. A bacterial assembly with less than 95 percent BUSCO completeness should trigger immediate investigation. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes how the tool assesses both eukaryotic and prokaryotic species, and the interpretation guidance differs by domain. Contiguity thresholds depend on the expected chromosome structure of your species. If you know the approximate chromosome number and size for your organism, you can set a target N50 that approaches the size of the largest chromosome. For the European mink, the N50 of 154.1 megabases approaches the size of the largest chromosome in the 2586.3 megabase genome. For a species with 20 chromosomes averaging 50 megabases each, an N50 below 10 megabases indicates that the assembly is far from chromosome-level resolution. Record your thresholds in your laboratory notebook or project documentation before running the assessment. This practice prevents post hoc rationalization of poor results and makes your quality standards transparent to collaborators and reviewers. ### Step 2: Run the Three-Layer Assessment The decision framework uses three layers of evidence. Layer one is contiguity, measured by N50 and L50. Layer two is gene-space completeness, measured by BUSCO. Layer three is read mapping consistency, which validates the physical structure of the assembly. For layer one, run QUAST or a similar tool to compute N50, L50, and related statistics. Report contig N50 and scaffold N50 separately. The [AquaaG pipeline](https://doi.org/10.1016/j.mex.2026.103955) integrates QUAST into an automated workflow that produces assembly quality reports, which reduces the risk of inconsistent metric calculation across multiple assemblies. For layer two, run BUSCO with the appropriate lineage dataset. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes the auto-lineage workflow that detects the appropriate dataset automatically, which is useful when you are uncertain about the taxonomic placement of your sample. Record the complete, duplicated, fragmented, and missing percentages separately. Do not collapse these into a single completeness score, because the distribution matters for diagnosis. For layer three, map the raw sequencing reads back to the assembly. Calculate the percentage of reads that map, the coverage uniformity across the assembly, and the number of regions with zero coverage. This step detects problems that N50 and BUSCO cannot see, such as collapsed repeats, chimeric joins, and underassembled regions. The [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible workflows for read mapping and coverage analysis that can be applied to your own data. ### Step 3: Apply the Triage Matrix Once you have all three layers of evidence, use the triage matrix to classify your assembly into one of four categories. Each category has specific action items. **Category A: High contiguity, high completeness, consistent read mapping.** This assembly is ready for downstream analysis. The N50 approaches the expected chromosome size, BUSCO completeness exceeds your threshold, and read mapping shows uniform coverage. Proceed with your intended analysis, but continue to monitor for local misassemblies that global metrics cannot detect. **Category B: High contiguity, low completeness.** This assembly is physically large but missing genomic regions. The [root-lesion nematode assembly](https://pubmed.ncbi.nlm.nih.gov/37108472) exemplifies this pattern with an N50 of 1.72 megabases but only 65.4 percent BUSCO completeness and 32.8 percent missing BUSCOs. The missing regions may be repetitive, GC-rich, or underrepresented in the sequencing data. Action items include examining the coverage distribution to identify zero-coverage regions, checking whether the missing BUSCOs are concentrated in specific genomic regions, and considering additional sequencing with a different library preparation to capture the missing regions. **Category C: Low contiguity, high completeness.** This assembly contains the expected gene space but is fragmented into many small pieces. The assembly may be usable for gene-based analyses such as annotation or phylogenomics, but it will be challenging for structural variant calling or chromosome-level comparisons. Action items include attempting scaffolding with Hi-C or optical mapping data, adjusting assembler parameters to produce longer contigs, or accepting the fragmentation and restricting downstream analyses to gene-centric questions. **Category D: Low contiguity, low completeness.** This assembly is fundamentally inadequate for most analyses. The [root-lesion nematode assembly](https://pubmed.ncbi.nlm.nih.gov/37108472) falls into this category when compared with the quality expected for a reference genome. Action items include verifying that the BUSCO lineage dataset is correct, checking read quality and adapter contamination, and considering a different assembly strategy or additional sequencing. The triage matrix also incorporates read mapping consistency as a modifier. If read mapping reveals large regions of zero coverage or highly uneven coverage, this information overrides the contiguity and completeness categories. An assembly with high N50 and high BUSCO completeness but large zero-coverage regions has a structural problem that the global metrics miss. ### Step 4: Document the Decision and Its Rationale After classifying your assembly, document the decision and the evidence that supports it. This documentation serves several purposes. It provides a record for your own troubleshooting if downstream analyses fail. It communicates quality limitations to collaborators who may use the assembly. It supports manuscript preparation by providing a transparent account of assembly quality. Create a decision record that includes the following fields: assembly identifier, date of assessment, software versions for all tools, BUSCO lineage dataset and version, contig N50, scaffold N50, L50, BUSCO completeness percentages, read mapping percentage, coverage uniformity summary, classification category, and the action taken. The [nf-core documentation](https://nf-co.re/docs) emphasizes reproducibility standards that include detailed logging of parameters and versions, which supports this type of decision documentation. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) provides an example of transparent quality documentation. The authors report two haplotypes with total lengths of 868.84 and 869.52 megabases, with most of haplotype 1 scaffolded into 30 chromosomal pseudomolecules. This level of detail allows other researchers to understand the assembly structure and to make informed decisions about whether to use it for their analyses. ### Step 5: Reassess After Any Improvement Attempt Assembly improvement is an iterative process. After you take action based on the triage classification, rerun the full three-layer assessment to determine whether the action resolved the problem. Do not assume that adding more sequencing data or switching assemblers will automatically improve quality. Measure the improvement directly. For example, if you add Hi-C data to scaffold a fragmented assembly, rerun QUAST to compute the new scaffold N50 and check whether the scaffolding produced chromosome-scale pseudomolecules. If you add sequencing data to improve BUSCO completeness, rerun BUSCO with the same lineage dataset and version to ensure comparability. The [European mink assembly](https://pubmed.ncbi.nlm.nih.gov/37834264) demonstrates the value of iterative improvement, with Hi-C data clustering more than 99.9 percent of the assembly into 20 chromosomal pseudomolecules. Keep a running log of each improvement attempt, the metrics before and after, and the outcome. This log becomes a valuable troubleshooting resource for future projects and provides evidence for the effectiveness of different assembly strategies. ### Common Failure Patterns in the Decision Framework Several failure patterns recur when researchers apply this decision framework. Recognizing these patterns helps you avoid common mistakes. **Pattern 1: Thresholds set after seeing results.** Researchers who set thresholds after viewing their BUSCO and N50 values tend to rationalize poor quality. A 70 percent BUSCO completeness becomes acceptable if the researcher decides that the species is difficult to assemble. This practice undermines the value of quality assessment. Set thresholds before running the assessment and document them. **Pattern 2: Comparing metrics across incompatible contexts.** N50 values are not directly comparable across species with different genome sizes and repeat content. A bacterial genome with an N50 of 4 megabases may be a single complete chromosome, while a plant genome with the same N50 may represent only a small fraction of the total sequence. Always compare assemblies within a taxonomic context. **Pattern 3: Ignoring read mapping evidence.** Some researchers rely solely on N50 and BUSCO and skip the read mapping layer. This is a significant omission because read mapping detects structural problems that global metrics cannot see. A chimeric contig can inflate N50, and a collapsed repeat can produce false BUSCO duplication. Read mapping provides the evidence needed to validate the physical structure of the assembly. **Pattern 4: Using the wrong BUSCO lineage dataset.** The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) emphasizes the importance of selecting the appropriate lineage dataset. Using a vertebrate dataset on a fungal assembly produces misleading results because many vertebrate-specific genes will be missing. The auto-lineage workflow can detect the appropriate dataset automatically, but you should verify that the selected lineage matches your species. **Pattern 5: Treating BUSCO duplication as always problematic.** High BUSCO duplication can indicate a true recent whole-genome duplication or a haplotype that was not collapsed during assembly. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) explicitly reports two haplotypes, which is appropriate for a heterozygous individual. Before treating duplication as an error, consider whether your species has undergone recent whole-genome duplication or whether you intended to produce a haplotype-resolved assembly. ### Records and Measurements for the Decision Framework The decision framework requires systematic record keeping to be effective. Create a standardized quality assessment form that you use for every assembly. The form should include the following fields: assembly identifier, species, sequencing platform, assembler and version, QUAST version, BUSCO version, BUSCO lineage dataset, contig N50, scaffold N50, L50, total assembly length, number of contigs and scaffolds, GC content, BUSCO complete percentage, BUSCO duplicated percentage, BUSCO fragmented percentage, BUSCO missing percentage, read mapping percentage, coverage uniformity summary, classification category, and action taken. The [European mink assembly](https://pubmed.ncbi.nlm.nih.gov/37834264) reports a genome size of 2586.3 megabases across 25 scaffolds, with an N50 of 154.1 megabases and a GC content of 41.9 percent. The [Parastichtis suspecta assembly](https://doi.org/10.12688/wellcomeopenres.26587.1) reports two haplotypes with total lengths of 868.84 and 869.52 megabases. These records demonstrate the level of detail that supports meaningful quality assessment and comparison. Store your quality assessment records in a version-controlled format that allows you to track changes over time. The [Carpentries lessons](https://carpentries.org/lessons) cover version control with Git, which is directly applicable to managing assembly quality records. The [Bioconductor project](https://bioconductor.org/) offers R packages for data management and visualization that can help you analyze quality metrics across multiple assemblies. ### Professional Escalation Criteria The decision framework includes clear escalation criteria that indicate when you should seek additional resources or expertise. Escalate to additional sequencing, a different assembly strategy, or collaboration with a genome assembly specialist when you observe any of the following conditions. First, BUSCO completeness is below 80 percent for a eukaryotic genome and the missing BUSCOs are distributed across the genome instead of concentrated in a few regions. This pattern suggests that the sequencing data are insufficient to cover the full gene space. The [root-lesion nematode assembly](https://pubmed.ncbi.nlm.nih.gov/37108472) with 32.8 percent missing BUSCOs exemplifies this condition, and the authors noted that the assembly was generated from only 500 nematodes using an ultra-low DNA input workflow. Second, read mapping reveals large regions of zero coverage that are not explained by known repeat content or assembly gaps. This pattern indicates that the assembly is missing genomic regions that were not sequenced or were not assembled. Additional sequencing with a different library preparation may be necessary to capture these regions. Third, the N50 is more than 10 times smaller than the expected chromosome size for your species and you need chromosome-level resolution for your analysis. This condition indicates that the assembly is far from the contiguity required for structural variant calling or chromosome-level comparative genomics. Hi-C scaffolding or additional long-read sequencing may be necessary. Fourth, BUSCO duplication is above 10 percent and you intended to produce a haploid assembly. This condition indicates that haplotypes were not collapsed during assembly. Haplotype purging or a different assembly strategy may be necessary to produce a haploid representation of the genome. Before escalating, verify that your BUSCO lineage dataset is correct and that your quality assessment tools are up to date. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes the auto-lineage workflow that can help you select the appropriate dataset. Also verify that your assembly software and parameters are appropriate for your data type. Different assemblers perform differently on different genomes, and switching assemblers may resolve quality problems without additional sequencing. When you escalate, document the specific metrics that triggered the escalation and the steps you took to verify the problem. This documentation is valuable for troubleshooting and for communicating with collaborators or reviewers. The [nf-core documentation](https://nf-co.re/docs) emphasizes reproducibility standards that include detailed logging of parameters and versions, which supports this type of troubleshooting. ### Integrating the Decision Framework with Reproducible Workflows The decision framework is most effective when integrated into a reproducible workflow. The [AquaaG pipeline](https://doi.org/10.1016/j.mex.2026.103955) demonstrates how assembly quality assessment can be automated and integrated with annotation. The pipeline retrieves assemblies from NCBI, runs QUAST for assembly quality assessment, runs BUSCO for gene-space completeness evaluation, and then proceeds to annotation. This type of automated workflow reduces the risk of inconsistent quality assessment across multiple assemblies. For researchers who prefer to build their own workflows, the [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible tutorials for assembly quality assessment that can be adapted to your specific needs. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide structured learning pathways for bioinformatics analysis, including assembly and quality assessment modules. The [nf-core documentation](https://nf-co.re/docs) describes community pipelines that follow reproducibility standards, which is valuable when you need to share your analysis with collaborators or reviewers. The [Bioconductor project](https://bioconductor.org/) offers R packages for genomic analysis that can complement assembly quality assessment. These packages can help you visualize quality metrics, compare assemblies, and integrate quality information into downstream analyses. The [Carpentries lessons](https://carpentries.org/lessons) cover foundational computing skills such as shell scripting and data management, which are necessary for running assembly quality tools efficiently, especially when working with large eukaryotic genomes that require substantial computational resources. The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) provide access to thousands of assemblies with reported quality metrics. Browsing these records can help you understand the range of quality values across taxa and set realistic thresholds for your own species. When you deposit your assembly in NCBI, include the quality metrics in the assembly metadata to help other researchers assess the quality of your assembly before using it. ## Frequently Asked Questions ### What is the difference between contig N50 and scaffold N50? Contig N50 is calculated from the unjoined contigs produced by the assembler. Scaffold N50 is calculated after scaffolding, which joins contigs with gaps to produce longer sequences. Scaffold N50 is always higher than or equal to contig N50 because scaffolding increases the length of the largest sequences. When comparing assemblies, check which N50 is reported. Reporting scaffold N50 without specifying that gaps are present can inflate the apparent quality. ### Why does my BUSCO completeness differ between runs? BUSCO results can vary between runs for several reasons. Different versions of BUSCO may use different ortholog datasets or search parameters. The choice of lineage dataset affects the results, and using the wrong lineage will produce misleading completeness scores. The quality of the assembly itself does not change between runs, but the search sensitivity can. Always record the BUSCO version and lineage dataset used so that you can compare results across runs. ### Can a high N50 assembly still be missing genes? Yes. N50 measures contiguity, not gene content. An assembly can have a high N50 and still be missing large genomic regions, including regions that contain conserved genes. This can happen when repetitive or GC-rich regions are difficult to sequence or assemble. BUSCO completeness is the metric that detects missing genes, and you should always run BUSCO in addition to computing N50. ### What does a high BUSCO duplication percentage mean? A high BUSCO duplication percentage means that many conserved single-copy genes are present in more than one copy in your assembly. This can indicate a recent whole-genome duplication in your species, or it can indicate that haplotypes were not collapsed during assembly. If you intended to produce a haploid assembly, high duplication suggests that you need to apply haplotype collapsing or purging steps. ### How do I choose the correct BUSCO lineage dataset? The BUSCO tool includes an auto-lineage workflow that can detect the appropriate dataset based on the content of your assembly. You can also select the lineage manually based on the taxonomic classification of your organism. Using the wrong lineage will produce misleading results, so verify that the selected lineage matches your species. The [BUSCO protocol](https://pubmed.ncbi.nlm.nih.gov/34936221) describes the available workflows and how to interpret the results. ### Can I compare N50 values across different species? N50 values are most meaningful when compared within the same species or between species with similar genome sizes and repeat content. Species with large genomes or high repeat content will have lower N50 values than species with small genomes or low repeat content, even when both assemblies are of similar quality. Always consider the genome biology of your species when interpreting N50. ### What additional metrics should I report alongside N50 and BUSCO? Report the total assembly length, number of contigs and scaffolds, L50, GC content, and the BUSCO lineage dataset and version. If you have read mapping data, report the percentage of reads that map to the assembly and the coverage uniformity. If you have Hi-C data, report the number of chromosome-scale scaffolds and the percentage of the assembly that is scaffolded into chromosomes. These additional metrics provide context for interpreting N50 and BUSCO. ### When should I consider generating more sequencing data? Consider generating more sequencing data when BUSCO completeness is below 90 percent for a eukaryotic genome, when read mapping reveals large regions of zero coverage, or when the assembly is too fragmented for your intended analysis. Before generating more data, verify that your assembly strategy and parameters are appropriate. Switching assemblers or adjusting parameters may resolve quality problems without additional sequencing. ## Related Bioinformatics Guides - [Evaluating Genome Assembly Quality: Metrics and Tools](/knowledge/bioinformatics/evaluating-genome-assembly-quality-metrics-and-tools) - [Hybrid Genome Assembly: Combining Short and Long Reads for Better Results](/knowledge/bioinformatics/hybrid-genome-assembly-combining-short-and-long-reads-for-better-results) - [How to Interpret Gene Set Enrichment Analysis Results](/knowledge/bioinformatics/how-to-interpret-gene-set-enrichment-analysis-results) - [De Novo Genome Assembly with Long Reads: A Practical Workflow](/knowledge/bioinformatics/de-novo-genome-assembly-with-long-reads-a-practical-workflow) - [Single-Cell RNA Sequencing Quality Control: A Practical Guide to Filtering and Metrics](/knowledge/bioinformatics/single-cell-rna-sequencing-quality-control-a-practical-guide-to-filtering-and-metrics) ## References and Further Reading - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Bioconductor](https://bioconductor.org/). Bioconductor Project. - [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project. - [nf-core Documentation](https://nf-co.re/docs). nf-core. - [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries. - [BUSCO: Assessing Genomic Data Quality and Beyond.](https://pubmed.ncbi.nlm.nih.gov/34936221). Current protocols, 2021. - [First Draft Genome Assembly of Root-Lesion Nematode Pratylenchus scribneri Generated Using Long-Read Sequencing.](https://pubmed.ncbi.nlm.nih.gov/37108472). International journal of molecular sciences, 2023. - [Prioritizing Endangered Species in Genome Sequencing: Conservation Genomics in Action with the First Platinum-Standard Reference-Quality Genome of the Critically Endangered European Mink Mustela lutreola L., 1761.](https://pubmed.ncbi.nlm.nih.gov/37834264). International journal of molecular sciences, 2023. - [The genome sequence of the Suspected, <,i>,Parastichtis suspecta<,/i>, (Hübner, 1809) (Lepidoptera: Noctuidae).](https://doi.org/10.12688/wellcomeopenres.26587.1). 2026. - [AquaaG: A comprehensive pipeline for quality assessment and annotation of genomes.](https://doi.org/10.1016/j.mex.2026.103955). 2026. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.