Evaluating Reference-Guided Assemblies: Metrics and Tools for Assessing Accuracy and Completeness

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

Key Takeaways

• Reference-guided assembly quality is assessed by genome coverage (proportion of reference covered), base-level accuracy (nucleotide identity to reference), structural variant recall (detection of large insertions/deletions/inversions), contig continuity (N50/L50 metrics), and gene completeness (presence of conserved orthologs via BUSCO).
• QUAST is a primary tool for evaluating genome coverage, base-level accuracy, and contig continuity (N50), while BUSCO is essential for quantifying gene completeness by identifying conserved single-copy orthologs.
• Structural variant recall, crucial for identifying large genomic rearrangements, is assessed using specialized tools like Assemblytics and Inspector, which analyze alignments to detect insertions, deletions, inversions, and duplications.
• The choice of evaluation metrics and tools is dictated by the downstream application; variant discovery prioritizes base-level accuracy and coverage (QUAST), structural variant analysis prioritizes recall (Assemblytics, Inspector), and gene annotation prioritizes completeness (BUSCO).
• Common failure patterns include over-reliance on a potentially incomplete reference genome, overlooking significant structural variants, using an inappropriate reference organism, and insufficient sequencing depth, all of which can lead to inaccurate biological conclusions.
• Interpretation of assembly metrics requires context, considering organism complexity (e.g., repetitive regions in eukaryotes) and sequencing technology, and validation with independent data or methods is critical for confirming assembly accuracy.

Reference-guided assembly uses an existing genome sequence as a framework to organize and orient sequencing reads from a new sample. The central problem researchers face is determining how well their reference-guided assembly represents the true genome of the organism under study. This article explains the core metrics used to evaluate assembly quality, describes practical tools for computing these metrics, and provides a workflow for interpreting results in the context of real research decisions.

Scope and Reader Context

This article addresses researchers, biology students, and laboratory professionals who have generated sequencing data and are using a reference genome to produce an assembly. The focus is on evaluating the output of reference-guided assembly pipelines, not on the mechanics of running aligners or assemblers. You will learn which metrics matter, how to compute them with established tools, and how to interpret the numbers in terms of biological confidence. The guidance applies to bacterial genomes, eukaryotic genomes, organellar genomes, and transcriptome assemblies where a reference is available. The practical outcome is the ability to determine whether your assembly is accurate enough for downstream applications such as variant calling, comparative genomics, or gene expression analysis.

At a Glance: Key Metrics and Tools for Assembly Evaluation

The table below summarizes the primary metrics used to assess reference-guided assemblies and the tools commonly used to compute them. Each metric answers a different question about assembly quality, and no single metric provides a complete picture.

Understanding Reference-Guided Assembly

The Role of the Reference Genome

A reference genome serves as a template for organizing sequencing reads. When a high-quality reference exists for a species or a closely related taxon, researchers can map reads to that reference and generate a consensus sequence for their sample. This approach is computationally efficient compared to de novo assembly and can produce highly accurate results when the sample is genetically similar to the reference. The quality of the reference genome directly influences the quality of the guided assembly. A fragmented or error-prone reference will propagate its limitations into the new assembly.

The National Center for Biotechnology Information maintains curated reference genomes and provides search systems for identifying appropriate references for your organism of interest. Researchers should verify that the chosen reference is the most current version and that its own assembly quality has been documented. The NCBI data resources include assembly statistics and annotation information that can inform this decision.

Reference-Guided versus De Novo Approaches

Reference-guided assembly and de novo assembly answer different questions and have different strengths. De novo assembly builds a genome from scratch, which is necessary when no reference exists, such as in non-model organisms. Long-read sequencing technologies have improved de novo assembly substantially because extended DNA sequences can span complex repetitive regions. However, de novo assembly remains computationally demanding and can produce fragmented results in repetitive or highly heterozygous genomes.

Reference-guided assembly leverages existing knowledge to resolve ambiguities that would challenge de novo methods. This approach is particularly useful when the goal is to identify variants in a sample relative to a known genome, such as in outbreak investigations or population studies. A study of Cyclospora cayetanensis, a parasite associated with foodborne outbreaks, demonstrated that a curated mitochondrial reference genome enabled the construction of new mitochondrial genomes from metagenomic reads. The researchers identified nucleotide variants in new and publicly available genomes by comparing them to the reference. This workflow proved useful for strain-level subtyping during outbreak investigations, a task that conventional epidemiological methods could not accomplish reliably.

The choice between reference-guided and de novo assembly depends on the research question. If you need to detect structural variants or characterize novel genomic content, de novo assembly may be more appropriate. If your goal is to identify single nucleotide polymorphisms or small indels in a well-characterized species, reference-guided assembly offers efficiency and accuracy.

Core Metrics for Assembly Evaluation

Genome Coverage

Genome coverage measures the proportion of the reference genome that is represented in your assembly. This metric is expressed as a percentage, where 100 percent indicates that every base of the reference is covered by at least one assembled contig. Coverage below 100 percent means that some regions of the reference are missing from your assembly.

Low coverage can result from several causes. Sequencing depth may be insufficient in certain regions, particularly those with extreme GC content or repetitive sequences. The assembly pipeline may have filtered out reads that map ambiguously. The sample may contain sequences that are absent from the reference, such as novel insertions or plasmid content in bacterial samples.

The jackfruit genome study provides a practical example of coverage assessment. The researchers generated a de novo assembly of 817.7 Mb and a reference-guided assembly of 843 Mb for the BARI Kanthal-3 variety. The difference in assembly size between the two approaches illustrates that reference-guided methods can recover additional sequence content by using the reference to organize reads that de novo assembly leaves unplaced.

Base-Level Accuracy

Base-level accuracy describes how closely the assembled sequence matches the true genome at individual nucleotide positions. In reference-guided assembly, this metric is often expressed as the identity between the assembled contigs and the reference genome. High identity indicates few substitution errors, while low identity suggests systematic errors in the assembly or genuine biological divergence between the sample and the reference.

It is important to distinguish between assembly errors and true biological variation. If your sample is from a different strain or individual than the reference, nucleotide differences may represent genuine polymorphisms instead of assembly mistakes. The Cyclospora study illustrates this distinction. The researchers identified nucleotide variants in newly assembled genomes by comparing them to the reference genome. These variants were biologically meaningful for strain typing, not artifacts of the assembly process.

Tools such as QUAST and Inspector compute base-level accuracy by aligning assembled contigs to the reference and counting mismatches and indels. The results are reported as error rates or quality values. A quality value of Q30 corresponds to an error rate of one per thousand bases, while Q40 corresponds to one error per ten thousand bases.

Structural Variant Recall

Structural variant recall measures the ability of your assembly to detect large-scale genomic changes, including insertions, deletions, inversions, and duplications. These variants are typically longer than 50 base pairs and can have substantial biological consequences. Reference-guided assembly can detect structural variants by identifying regions where the assembled sequence does not align collinearly with the reference.

Assemblytics is a tool specifically designed to detect and quantify structural variants from assembly alignments. It identifies insertions, deletions, and tandem expansions by analyzing the alignment of assembled contigs to a reference genome. The output includes the size distribution of variants and their genomic locations.

Inspector provides additional capabilities for structural variant detection. According to its protocol documentation, Inspector can detect both small-scale and large-scale structural errors in assemblies. It supports both reference-free and reference-guided evaluation, making it versatile for different analysis scenarios. Inspector also offers assembly error correction, which can improve the quality value of the original assembly.

The honey bee genome study demonstrates the importance of structural variant detection in repetitive regions. The researchers used long-read-based assemblies to resolve gaps in the reference genome, including 13 gaps, five unplaced scaffolds, and missing telomeres. The total length of the resolved gaps was 848,747 base pairs. This work showed that even high-quality reference genomes can have structural deficiencies in highly repetitive regions, and that comparative analysis of multiple assemblies can identify and correct these problems.

Contig Continuity

Contig continuity describes the degree to which the assembly is organized into long, contiguous sequences instead of many short fragments. Metrics such as N50 and L50 summarize continuity. N50 is the length of the shortest contig such that contigs of that length or longer account for at least half of the total assembly length. L50 is the number of contigs needed to reach that half-length threshold.

Higher N50 values indicate more contiguous assemblies. Reference-guided assembly generally improves continuity compared to de novo assembly because the reference provides a framework for ordering and orienting contigs. However, continuity can still be limited by sequencing gaps, repetitive regions, or insufficient coverage.

The long-read transcriptome assembly evaluation provides context for continuity in RNA sequencing applications. The study compared long-read de novo transcriptome assembly tools to a leading short-read assembler. The results confirmed that long reads generate longer assembled transcripts than short reads for reference-free analysis. However, the authors noted that limitations remain compared to reference-guided approaches, suggesting that reference-guided methods still offer advantages in continuity for transcriptome applications.

Gene Completeness

Gene completeness assesses whether the assembly contains expected conserved genes. The Benchmarking Universal Single-Copy Orthologs (BUSCO) tool is the standard method for this assessment. BUSCO searches the assembly for a set of single-copy orthologs that are expected to be present in the target taxonomic group. The results classify each gene as complete, fragmented, or missing.

The jackfruit genome study reported BUSCO results of 97.2 percent complete core genes, with 1.3 percent fragmented and 1.5 percent missing. These numbers indicate a high-quality assembly with minimal gene content loss. Gene completeness is particularly important for downstream applications such as annotation and comparative genomics, where missing genes could lead to incorrect biological conclusions.

Tools for Computing Assembly Metrics

QUAST

QUAST is a widely used tool for evaluating genome assemblies. It computes a range of metrics including N50, genome coverage, misassembly rates, and base-level accuracy. QUAST aligns assembled contigs to a reference genome and reports statistics that allow comparison between multiple assemblies. The tool is suitable for both reference-guided and de novo assembly evaluation.

To use QUAST effectively, you need a reference genome in FASTA format and your assembly in FASTA format. QUAST will generate a report containing the key metrics described above. The report can be generated in text, HTML, or PDF format, facilitating sharing with collaborators or inclusion in publications.

Inspector

Inspector is a more recent tool that offers several advantages for long-read assembly evaluation. According to its protocol documentation, Inspector supports both reference-free and reference-guided evaluation, detects both small- and large-scale structural errors, offers assembly error correction, and can perform haplotype-resolved assembly evaluation. These capabilities make Inspector particularly useful for complex genomes where structural errors are a concern.

The Inspector protocol describes four procedures demonstrating different applications for long-read assembly evaluation. The tool provides basic contig and alignment statistics as well as precise locations and types of structural errors. This level of detail allows researchers to identify specific problem regions in their assembly and take corrective action.

Assemblytics

Assemblytics specializes in structural variant detection from assembly alignments. It takes as input the alignment of assembled contigs to a reference genome and produces a detailed report of insertions, deletions, and tandem expansions. The tool is particularly useful for identifying structural variants that would be missed by short-read variant calling approaches.

Assemblytics is appropriate when your research question involves large-scale genomic rearrangements. For example, if you are studying a bacterial outbreak and need to identify genomic islands or phage insertions, Assemblytics can detect these variants from your reference-guided assembly.

BUSCO

BUSCO assesses gene completeness by searching for conserved single-copy orthologs. The tool requires a lineage-specific database, which is downloaded based on your organism of interest. BUSCO can be run on both genome assemblies and transcriptome assemblies, making it versatile for different research contexts.

The European Bioinformatics Institute provides training materials on bioinformatics data resources and practical analysis education. These resources can help researchers learn how to run BUSCO and interpret its output correctly.

Practical Workflow for Evaluating a Reference-Guided Assembly

Step 1: Verify Input Data Quality

Before evaluating your assembly, confirm that the input sequencing data are of sufficient quality. Check the sequencing depth, read length, and base quality scores. Low-quality input data will produce a poor assembly regardless of the evaluation tools used. The Galaxy Training Network provides accessible workflow training that covers quality control of sequencing data and other foundational analysis steps.

Step 2: Select an Appropriate Reference

Choose a reference genome that is closely related to your sample. The reference should be the most current version available and should have documented assembly quality. The NCBI data resources provide search systems for identifying reference genomes and accessing their associated metadata. If your sample is from a species without a high-quality reference, consider whether reference-guided assembly is appropriate or whether de novo assembly would be more suitable.

Step 3: Generate the Reference-Guided Assembly

Use your preferred alignment and assembly tools to generate the reference-guided assembly. Document the parameters used, including the aligner, the minimum mapping quality threshold, and any filtering steps. Reproducibility is essential for scientific rigor. The nf-core documentation describes community standards for pipeline usage and configuration that can help ensure your assembly workflow is reproducible.

Step 4: Compute Assembly Metrics

Run QUAST or Inspector to compute genome coverage, base-level accuracy, and contig continuity. Run BUSCO to assess gene completeness. If structural variant detection is relevant to your research question, run Assemblytics or use Inspector's structural error detection capabilities. Record all metrics in a structured format for comparison with other assemblies or for inclusion in publications.

Step 5: Interpret Results in Biological Context

Interpret the metrics in light of your research question and the expected characteristics of your organism. A bacterial genome assembly should achieve near-complete coverage and high base-level accuracy. A eukaryotic genome with substantial repetitive content may have lower continuity and more missing regions. The honey bee genome study illustrates that even high-quality assemblies can have gaps in extended repetitive regions, and that resolving these gaps may require specialized approaches such as ultra-long Nanopore sequencing.

Step 6: Document and Report

Document your evaluation results and the parameters used to generate them. Include the version numbers of all tools and the reference genome version. This documentation enables other researchers to reproduce your work and assess the reliability of your conclusions. The Carpentries lessons provide foundational training in data management and reproducible research practices that are applicable to assembly evaluation.

Records and Measurements for Assembly Evaluation

Maintaining an Assembly Evaluation Log

Keep a structured log of all assembly evaluation activities. The log should include the sample identifier, the reference genome version, the assembly tool and version, the evaluation tools and versions, and the date of analysis. Record all computed metrics, including genome coverage, N50, base-level accuracy, structural variant counts, and BUSCO results. This log serves as a permanent record that supports publication and facilitates troubleshooting if problems arise.

Comparing Assemblies Across Samples

When evaluating multiple assemblies, use consistent metrics and tools to enable direct comparison. QUAST can compare multiple assemblies in a single run, generating a combined report. This capability is useful when testing different assembly parameters or when comparing assemblies from different samples. The long-read transcriptome assembly study provides an example of systematic comparison across multiple tools and datasets, using simulated data and spike-in sequin transcripts where ground truth was known.

Tracking Changes After Polishing

Assembly polishing is the process of correcting errors in an assembly using additional sequencing data or computational methods. Inspector offers assembly error correction that can improve the quality value of the original assembly. After polishing, rerun the evaluation metrics to quantify the improvement. Record the before and after values to document the effect of polishing.

Common Failure Patterns in Reference-Guided Assembly

Over-Trusting the Reference

A common failure is assuming that the reference genome is correct and complete. The honey bee genome study demonstrated that even a chromosome-level reference assembly can have gaps, unplaced scaffolds, and missing telomeres. If your sample contains sequences that are absent from the reference, reference-guided assembly will not recover them. This limitation is particularly relevant for bacterial samples that may contain plasmids or phage sequences not present in the reference strain.

Ignoring Structural Variants

Focusing exclusively on base-level accuracy can cause researchers to miss important structural variants. The Cyclospora study highlighted that the absence of reference genomes limited the application of sequencing data for source tracking during outbreak investigations. Structural variants such as insertions and deletions can be biologically significant, and tools like Assemblytics and Inspector should be used to detect them.

Using an Inappropriate Reference

Selecting a reference that is too distantly related to your sample will produce poor results. The assembly will contain many mismatches that reflect biological divergence instead of assembly errors, making it difficult to distinguish true variants from artifacts. If the reference is too divergent, consider whether de novo assembly or a different reference would be more appropriate.

Insufficient Sequencing Depth

Low sequencing depth leads to incomplete coverage and fragmented assemblies. The long-read transcriptome assembly study evaluated datasets ranging from 6 to 60 million reads, demonstrating that depth affects assembly quality. If your assembly shows low coverage or poor continuity, increasing sequencing depth may resolve the problem.

Failure to Validate with Independent Data

Relying on a single evaluation method can miss errors. The honey bee study validated the corrected assembly by mapping PacBio reads and performing gene annotation assessment. Independent validation approaches provide confidence that the assembly is accurate and complete.

Limitations of Reference-Guided Assembly Evaluation

Reference Bias

Reference-guided assembly is inherently biased toward the reference genome. Sequences that are present in the sample but absent from the reference will be missed. This bias is particularly problematic for structural variant detection, where novel insertions cannot be identified if they are not in the reference. The jackfruit genome study illustrated this limitation by showing that the reference-guided approach yielded a different assembly size than the de novo approach.

Difficulty in Distinguishing Errors from Variation

When the sample is genetically divergent from the reference, it becomes difficult to distinguish assembly errors from true biological variation. This challenge is particularly acute for structural variants, where alignment artifacts can mimic genuine rearrangements. Tools like Inspector provide precise locations and types of structural errors, which helps in making this distinction.

Computational Requirements

Some evaluation tools, particularly those that align long reads to assemblies, require substantial computational resources. The Inspector protocol describes procedures for long-read assembly evaluation that may be computationally intensive. Researchers should ensure they have adequate computing infrastructure before running these analyses.

Incomplete Ground Truth

For most real samples, the true genome sequence is unknown. Evaluation metrics provide indirect evidence of assembly quality but cannot prove that the assembly is correct. The long-read transcriptome assembly study used simulated data and spike-in sequin transcripts where ground truth was known, but such validation is not possible for most real samples.

Quality and Reproducibility Controls

Version Control for Tools and References

Record the exact versions of all tools and reference genomes used in your analysis. Software updates can change algorithm behavior and produce different results. Version control enables reproducibility and facilitates troubleshooting when results differ between analyses. The nf-core documentation emphasizes the importance of reproducible workflow configuration and usage.

Containerization and Workflow Management

Use containerized tools and workflow management systems to ensure that analyses are reproducible across different computing environments. The nf-core community provides standards for pipeline usage and configuration that support reproducibility. The Galaxy Training Network offers accessible workflow training that covers reproducible analysis practices.

Independent Validation

Validate your assembly using independent methods. Map raw reads back to the assembly to confirm that reads align consistently. Compare your assembly to independently generated assemblies if available. The honey bee study used multiple long-read-based assemblies to validate improvements to the reference genome, demonstrating the value of comparative validation.

Safety and Regulatory Context

Data Management and Privacy

Genome assembly data may include sensitive information, particularly if the samples come from human subjects or agricultural animals. Ensure that your data management practices comply with applicable regulations and institutional policies. The NCBI data resources provide guidance on data submission and access that can inform your data management decisions.

Responsible Reporting of Assembly Quality

When reporting assembly quality in publications or databases, provide complete and accurate metrics. Do not selectively report metrics that make the assembly look better than it is. The scientific community relies on honest reporting to assess the reliability of genomic data. The EMBL-EBI training resources provide guidance on responsible data reporting and analysis practices.

Professional Escalation Criteria

When to Seek Expert Assistance

If your assembly evaluation reveals persistent problems that you cannot resolve, seek assistance from bioinformatics experts. Indicators that expert help is needed include:

• Genome coverage below 90 percent for a bacterial genome
• BUSCO completeness below 90 percent for a eukaryotic genome
• Structural variant counts that are implausibly high or low
• Inconsistent results between different evaluation tools
• Assembly quality that does not improve after polishing

Consulting Specialized Resources

The Galaxy Training Network provides tutorials on assembly evaluation and related topics. The EMBL-EBI training portal offers courses on bioinformatics data resources and analysis methods. The Carpentries lessons provide foundational training in computing and data skills that can help you troubleshoot assembly problems independently.

Decision Framework for Choosing Evaluation Tools Based on Assembly Purpose

Selecting the right evaluation strategy requires matching the assessment approach to the specific downstream application of your reference-guided assembly. A single evaluation pipeline does not serve all research questions equally, and the choice of metrics and tools should reflect what you plan to do with the assembly after validation. This section provides a structured decision framework that connects assembly purpose to evaluation priorities, tool selection, and interpretation thresholds.

Define the Primary Use Case Before Evaluation

The first decision point is clarifying what the assembly will be used for. This determination shapes which metrics deserve the most attention and which tools will provide the most relevant information. Four common use cases cover most reference-guided assembly projects.

Variant discovery for population studies. If the goal is identifying single nucleotide polymorphisms and small indels across many samples, base-level accuracy and genome coverage take priority. The Cyclospora cayetanensis study exemplifies this use case, where the researchers needed a curated mitochondrial reference genome to identify nucleotide variants for strain-level subtyping during foodborne outbreak investigations. For this purpose, QUAST provides the essential base-level statistics, and the evaluation should focus on minimizing false variant calls caused by assembly errors.

Structural variant characterization. When the research question involves large insertions, deletions, inversions, or duplications, structural variant recall becomes the primary metric. Assemblytics and Inspector are the appropriate tools because they detect large-scale genomic changes that base-level metrics miss. The honey bee genome study demonstrated the importance of structural evaluation by resolving 13 gaps, five unplaced scaffolds, and missing telomeres in the reference genome, totaling 848,747 base pairs of previously unresolved sequence.

Gene content and annotation projects. If the assembly will be used for gene prediction, comparative genomics, or functional annotation, gene completeness as measured by BUSCO is the critical metric. The jackfruit genome study reported 97.2 percent complete core genes with 1.3 percent fragmented and 1.5 percent missing, providing a benchmark for what constitutes a high-quality assembly for annotation purposes.

Transcriptome analysis. For RNA sequencing applications where the assembly represents expressed transcripts instead of genomic DNA, the evaluation priorities shift toward transcript continuity and isoform representation. The long-read transcriptome assembly evaluation confirmed that long reads generate longer assembled transcripts than short reads for reference-free analysis, though limitations remain compared to reference-guided approaches. This finding suggests that transcriptome assemblies require evaluation of transcript length distributions and isoform completeness in addition to standard genomic metrics.

Match Tools to the Evaluation Priority

Once the primary use case is defined, select the evaluation tools that directly address the priority metrics. The table below summarizes the recommended tool selection based on assembly purpose.

Establish Thresholds Based on Organism and Data Type

Interpretation thresholds should be calibrated to the organism being studied and the sequencing technology used. A bacterial genome assembly should achieve near-complete coverage and high base-level accuracy because bacterial genomes are compact and generally lack the complex repetitive structure found in eukaryotic genomes. The Cyclospora study demonstrated that a curated mitochondrial reference genome enabled high-confidence variant identification, but the researchers noted that the absence of reference genomes had previously limited the application of sequencing data for source tracking.

For eukaryotic genomes, lower thresholds are acceptable because of the presence of repetitive regions and higher heterozygosity. The jackfruit genome study reported a heterozygosity rate of 1.62 percent and an estimated genome size of 1.04 gigabase pairs. The reference-guided approach yielded 843 megabases of genome sequence compared to 817.7 megabases from de novo assembly, illustrating that reference-guided methods can recover additional sequence content in heterozygous genomes.

The honey bee genome study provides a cautionary example for repetitive genomes. The researchers found that even chromosome-level assemblies had 51 gaps, 160 unplaced or unlocalized scaffolds, and missing distal telomeres. The gaps were located in extended highly repetitive chromosomal regions, and comparative analysis suggested that PacBio-read-based assemblies failed in the same regions, particularly on chromosome 10. This finding indicates that repetitive genomes require specialized evaluation approaches and that thresholds for continuity metrics should be adjusted accordingly.

Apply the Decision Framework in Practice

The following stepwise procedure applies the decision framework to a real evaluation scenario.

Step 1: Document the assembly purpose. Write a one-sentence statement of what the assembly will be used for. This statement guides all subsequent decisions. For example, the purpose might be to identify single nucleotide polymorphisms in clinical isolates of a bacterial pathogen for outbreak tracking, or to characterize structural variants in a crop species for breeding applications.

Step 2: Select the primary evaluation tool. Based on the purpose statement, choose the tool that directly measures the priority metric. For variant discovery, run QUAST first. For structural variant analysis, run Assemblytics or Inspector. For gene annotation, run BUSCO. For transcriptome analysis, run QUAST with transcript-specific parameters.

Step 3: Run secondary evaluations. After the primary evaluation, run complementary tools to provide context. Even if structural variants are not the primary focus, running Inspector can identify assembly errors that might affect base-level accuracy measurements. The Inspector protocol describes procedures for detecting both small-scale and large-scale structural errors, which can reveal problems that QUAST alone would miss.

Step 4: Compare against organism-specific thresholds. Interpret the results using thresholds appropriate for the organism and data type. For bacterial genomes, expect genome coverage above 95 percent and base-level accuracy above 99.9 percent. For eukaryotic genomes, expect BUSCO completeness above 90 percent and N50 values appropriate for the genome size and complexity. The long-read transcriptome assembly study used datasets ranging from 6 to 60 million reads, demonstrating that depth affects assembly quality and should be considered when interpreting results.

Step 5: Document the decision rationale. Record why specific tools were selected and how the results were interpreted. This documentation supports reproducibility and provides context for collaborators or reviewers who may question the evaluation approach. The nf-core documentation emphasizes the importance of reproducible workflow configuration, and the Carpentries lessons provide foundational training in data management practices that support this documentation effort.

Record System for Evaluation Decisions

Maintain a structured record that captures the decision framework application for each assembly project. The record should include the assembly purpose statement, the tools selected and their versions, the thresholds applied, the metric values obtained, and the interpretation reached. This record system enables comparison across projects and facilitates troubleshooting when assemblies fail to meet expected quality standards.

The Galaxy Training Network provides accessible workflow training that covers reproducible analysis practices, including documentation standards. The EMBL-EBI training portal offers courses on bioinformatics data resources that can inform record-keeping practices. The Bioconductor project provides package documentation that supports reproducible genomic analysis workflows.

Troubleshooting Method for Failed Evaluations

When an assembly fails to meet the thresholds established by the decision framework, use a systematic troubleshooting method to identify the cause.

Check reference appropriateness first. The honey bee study demonstrated that reference genomes can have structural deficiencies that propagate into reference-guided assemblies. Verify that the chosen reference is the most current version and that its own assembly quality has been documented. The NCBI data resources provide assembly statistics and annotation information that can inform this verification.

Examine coverage distribution. Low genome coverage often results from sequencing depth being insufficient in certain regions, particularly those with extreme GC content or repetitive sequences. The jackfruit study reported a heterozygosity rate of 1.62 percent, which can complicate assembly in repetitive regions. If coverage is uneven, consider whether additional sequencing or different library preparation methods would resolve the problem.

Evaluate structural error locations. If Inspector identifies structural errors, examine whether they cluster in specific genomic regions. The honey bee study found that assembly failures concentrated in extended highly repetitive regions, especially on chromosome 10. This pattern suggests that the errors reflect genomic complexity instead of pipeline problems, and that resolving them may require specialized approaches such as ultra-long Nanopore sequencing.

Compare against independent assemblies. If available, compare your assembly to independently generated assemblies from the same sample or closely related samples. The honey bee study used multiple long-read-based assemblies to validate improvements to the reference genome. This comparative approach provides confidence that the assembly is accurate and that observed problems reflect genuine genomic features instead of pipeline artifacts.

Consider polishing and re-evaluation. Inspector offers assembly error correction that can improve the quality value of the original assembly. After polishing, rerun the evaluation metrics to quantify the improvement. Record the before and after values to document the effect of polishing and to determine whether the assembly now meets the thresholds established by the decision framework.

Comparison of Evaluation Strategies Across Assembly Types

The decision framework applies differently across assembly types, and understanding these differences improves evaluation accuracy.

Bacterial genomes. Bacterial assemblies are typically small and compact, allowing near-complete coverage and high base-level accuracy. The Cyclospora study demonstrated that a curated mitochondrial reference genome enabled high-confidence variant identification for outbreak tracking. For bacterial genomes, the decision framework prioritizes base-level accuracy and genome coverage, with structural variant analysis as a secondary consideration.

Eukaryotic genomes. Eukaryotic assemblies are larger and contain more repetitive sequence, requiring adjusted thresholds. The jackfruit study reported a genome size of 1.04 gigabase pairs with 1.62 percent heterozygosity, and the reference-guided assembly recovered 843 megabases. For eukaryotic genomes, the decision framework prioritizes gene completeness and contig continuity, with structural variant analysis focused on known repetitive regions.

Organellar genomes. Mitochondrial and chloroplast genomes are small and often circular, requiring specialized evaluation approaches. The Cyclospora study used a curated mitochondrial reference genome to build new mitochondrial genomes from metagenomic reads. For organellar genomes, the decision framework prioritizes base-level accuracy and complete circularization, with structural variant analysis focused on rearrangements that may affect gene order.

Transcriptome assemblies. Transcriptome assemblies represent expressed sequences instead of complete genomes, requiring different evaluation metrics. The long-read transcriptome assembly study evaluated tools across datasets ranging from 6 to 60 million reads and found that long reads generate longer assembled transcripts than short reads. For transcriptome assemblies, the decision framework prioritizes transcript continuity and isoform representation, with gene completeness as a secondary metric.

Professional Escalation Criteria for the Decision Framework

When the decision framework reveals persistent problems that cannot be resolved through troubleshooting, escalate to expert assistance. Indicators that expert help is needed include:

• Genome coverage below 90 percent for a bacterial genome despite adequate sequencing depth
• BUSCO completeness below 90 percent for a eukaryotic genome after polishing
• Structural variant counts that are implausibly high or low relative to the organism
• Inconsistent results between different evaluation tools that cannot be explained by tool differences
• Assembly quality that does not improve after multiple polishing rounds

The Galaxy Training Network provides tutorials on assembly evaluation and related topics. The EMBL-EBI training portal offers courses on bioinformatics data resources and analysis methods. The Carpentries lessons provide foundational training in computing and data skills that can help troubleshoot assembly problems independently before seeking expert assistance.

Integrating the Decision Framework with Reproducibility Practices

The decision framework should be integrated with broader reproducibility practices to ensure that evaluation results are trustworthy and comparable across projects. Record the assembly purpose, tool versions, reference genome version, thresholds applied, and metric values obtained. The nf-core documentation describes community standards for pipeline usage and configuration that support reproducible workflows. The Bioconductor project provides package documentation that supports reproducible genomic analysis.

The Galaxy Training Network offers accessible workflow training that covers reproducible analysis practices. The EMBL-EBI training portal provides courses on data-resource training and practical analysis education. These resources support the documentation and reproducibility practices that make the decision framework effective across projects and research groups.

Frequently Asked Questions

What is the difference between genome coverage and sequencing depth?

Genome coverage measures the proportion of the reference genome represented in the assembly. Sequencing depth measures how many times each base was read on average. A genome can have high sequencing depth but low genome coverage if reads are concentrated in certain regions and absent from others. Both metrics are important for assessing assembly quality.

How do I choose between QUAST and Inspector for assembly evaluation?

QUAST is a general-purpose assembly evaluation tool that computes standard metrics such as N50, genome coverage, and base-level accuracy. Inspector offers additional capabilities including structural error detection, assembly error correction, and haplotype-resolved evaluation. If your research involves long-read assemblies or complex genomes with structural variation, Inspector provides more detailed information. For routine evaluation of bacterial or simple eukaryotic assemblies, QUAST may be sufficient.

Can reference-guided assembly detect novel sequences that are not in the reference?

Reference-guided assembly cannot recover sequences that are absent from the reference genome. Reads that do not map to the reference are typically discarded or left unassembled. If you need to detect novel sequences such as plasmids, phage insertions, or genomic islands, consider supplementing reference-guided assembly with de novo assembly of unmapped reads.

What BUSCO completeness score should I expect for a good assembly?

A good assembly should have BUSCO completeness above 90 percent for most organisms. The jackfruit genome study reported 97.2 percent complete core genes, which is considered high quality. Scores below 90 percent suggest that the assembly is missing substantial gene content and may require additional sequencing or assembly improvement.

How does the choice of reference genome affect assembly quality?

The reference genome directly influences assembly quality. A reference that is closely related to your sample will produce a more accurate assembly with fewer mismatches. A reference that is distantly related will introduce many mismatches that are difficult to distinguish from true variants. Always verify that the reference genome is the most current version and has documented assembly quality.

What should I do if my assembly has low structural variant recall?

Low structural variant recall may indicate that your assembly is missing large insertions or deletions. This problem can occur if the reference genome lacks sequences present in your sample or if the assembly pipeline filters out reads that map ambiguously. Consider using Inspector to detect structural errors and Assemblytics to quantify structural variants. Increasing sequencing depth or using longer reads may also improve structural variant detection.

Is reference-guided assembly appropriate for transcriptome analysis?

Reference-guided assembly can be used for transcriptome analysis when a high-quality reference genome is available. The long-read transcriptome assembly study found that reference-guided approaches still offer advantages over de novo methods for transcriptome assembly. However, the study also noted limitations in reference-free analysis compared to reference-guided approaches, suggesting that the choice depends on the availability and quality of the reference.

How do I report assembly evaluation metrics in a publication?

Report all computed metrics including genome coverage, N50, base-level accuracy, structural variant counts, and BUSCO results. Include the versions of all tools and the reference genome used. Describe the parameters used for assembly and evaluation. This information allows other researchers to reproduce your work and assess the reliability of your conclusions.

Related Bioinformatics Guides

• Evaluating Genome Assembly Quality: Metrics and Tools
• Evaluating Metagenomic Assembly Tools: A Benchmarking Framework for Short-Read and Long-Read Data
• Genomic Data Analysis Tools: A Comparative Guide for Researchers
• RNA-Seq Quality Control: Essential Checks and Tools
• De Novo Genome Assembly with Long Reads: A Practical Workflow

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.
• A hybrid reference-guided de novo assembly approach for generating Cyclospora mitochondrion genomes.. Gut pathogens, 2018.
• A comprehensive evaluation of long-read de novo transcriptome assembly.. Genome biology, 2026.
• A detailed guide to assessing genome assembly based on long-read sequencing data using Inspector.. Nature protocols, 2025.
• Improved Apis mellifera reference genome based on the alternative long-read-based assemblies.. G3 (Bethesda, Md.), 2021.
• Whole-genome sequencing of a year-round fruiting jackfruit (Artocarpus heterophyllus Lam.) reveals high levels of single nucleotide variation.. Frontiers in plant science, 2022.

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