From Raw Reads to Finished Genomes

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

Quick Answer

• Use Unicycler for hybrid bacterial genome assembly when you have both Illumina short reads and Oxford Nanopore or PacBio long reads, because it produces accurate, complete assemblies even with low long-read depth.
• Start with a short-read-only assembly using SPAdes or SKESA to assess data quality, then add long reads through Unicycler's semi-global alignment approach to resolve repetitive regions.
• Hybrid assembly cannot fully resolve every plasmid or repeated element, and assembly quality depends heavily on input read quality, coverage depth, and the specific bacterial genome structure.

At a Glance

Understanding Hybrid Assembly for Bacterial Genomes

The Problem with Short Reads Alone

Illumina sequencing generates accurate but short reads, typically 150 to 300 base pairs in length. These reads provide high per-base accuracy, but the assembly process struggles when the genome contains repetitive elements longer than the read length. Bacterial genomes are full of such repeats, including ribosomal RNA operons, insertion sequences, and phage-related elements. When short reads cannot span these repeats, the assembler produces fragmented contigs with ambiguous connections.

The practical consequence is that short-read-only assemblies of bacterial genomes often contain dozens or hundreds of contigs instead of a complete chromosome. This fragmentation matters for downstream analysis. For example, plasmid content is difficult to analyze from draft assemblies because plasmids frequently contain repetitive elements and mobile genetic elements that break assembly continuity. The MOB-suite developers noted that analyzing plasmid content from short-read draft assemblies remains difficult due to incomplete assembly of plasmids, and bacterial isolates can contain any number of plasmids with assembly complicated by repetitive elements.

The Problem with Long Reads Alone

Pacific Biosciences and Oxford Nanopore Technologies platforms generate long reads that can span repetitive regions and produce complete genome assemblies. However, these platforms have higher per-base error rates than Illumina, and the sequencing is more expensive. A complete assembly from long reads alone requires substantial sequencing depth to overcome the error rate through consensus, which increases cost.

The complementary strengths of the two platforms create a clear rationale for hybrid assembly. Short reads provide accuracy, while long reads provide structural resolving power. Unicycler was designed specifically to leverage both types of data, building an initial assembly graph from short reads using SPAdes and then simplifying the graph using information from both short and long reads.

What Unicycler Does Differently

Unicycler uses a novel semi-global aligner to align long reads to the assembly graph. This approach allows the tool to use long-read information to resolve ambiguous paths through the graph without requiring the long reads to be error-free. Tests on synthetic and real reads show that Unicycler can assemble larger contigs with fewer misassemblies than other hybrid assemblers, even when long-read depth and accuracy are low.

The practical implication is that you do not need extremely deep long-read coverage to achieve a high-quality hybrid assembly. This makes hybrid assembly cost-effective for laboratories that already have Illumina infrastructure and can add a modest amount of long-read sequencing.

Preparing Input Data for Hybrid Assembly

Assessing Short-Read Quality

Before running any assembly, you need to evaluate the quality of your Illumina reads. Poor-quality input data produces poor assemblies regardless of the assembler you choose. Start by examining per-base quality scores, adapter contamination, and GC bias in your read set.

For bacterial genomics, you typically want paired-end reads with at least 50-fold coverage of the expected genome size. Lower coverage can still produce useful assemblies, but you should expect more fragmentation. The SKESA assembler was designed for assembling microbial genomes sequenced using Illumina and handles low-level contamination in reads, which makes it a useful tool for quality assessment and initial assembly attempts.

Evaluating Long-Read Quality and Depth

Long-read platforms produce reads with variable quality. Oxford Nanopore reads have improved substantially in recent years, but you should still check read length distributions and estimated error rates before assembly. For hybrid assembly with Unicycler, you do not need the same depth as you would for a long-read-only assembly, but you need enough long-read coverage to resolve the repetitive regions that break short-read contiguity.

A reasonable starting point is 20 to 50-fold coverage from long reads, but the optimal depth depends on your genome's repeat structure. Genomes with large or numerous repeats require more long-read coverage. You can assess whether you have sufficient long-read depth by examining the assembly graph complexity after the initial short-read assembly step.

Read Trimming and Contamination Control

Trim adapters and low-quality bases from both short and long reads before assembly. For Illumina reads, tools like Trimmomatic or fastp are standard. For Oxford Nanopore reads, Porechop removes adapter sequences, and you should also filter reads below a minimum length threshold.

Contamination is a particular concern in bacterial genomics. SKESA was shown to handle low-level contamination in reads while maintaining high sequence quality and contiguity. However, you should still check for contamination before assembly by examining k-mer distributions or by mapping reads to your expected species. Contamination from other bacteria or from the sequencing facility can produce chimeric assemblies or inflate genome size estimates.

Running Unicycler for Hybrid Assembly

Basic Command Structure

Unicycler accepts short reads in FASTQ format and long reads in FASTQ or FASTA format. The basic command structure is:

unicycler -1 short_reads_R1.fastq.gz -2 short_reads_R2.fastq.gz -l long_reads.fastq.gz -o output_directory

For Oxford Nanopore reads, you can add the --nanopore flag to optimize settings for that platform. For PacBio reads, use --pacbio. These flags adjust alignment parameters to account for platform-specific error profiles.

Parameter Selection

Unicycler has several parameters that affect assembly quality. The most important are the minimum contig length for the short-read assembly step and the minimum depth for long-read alignment. Default parameters work well for most bacterial genomes, but you may need to adjust them for unusual genomes.

If your genome has very high GC content, you may need to adjust the k-mer sizes used in the SPAdes step. Unicycler allows you to specify k-mer sizes with the --kmer parameter. For high-GC organisms, smaller k-mers often perform better because they are less likely to fall in repetitive GC-rich regions.

For genomes with very high coverage, you can use the --keep parameter to retain intermediate files for troubleshooting. This is useful when an assembly fails or produces unexpected results, because you can examine the assembly graph at each stage.

The Assembly Pipeline

Unicycler operates in three stages. First, it builds an assembly graph from short reads using SPAdes. Second, it aligns long reads to this graph using the semi-global aligner. Third, it uses the long-read alignments to resolve ambiguous paths and produce the final assembly.

The output includes the final assembly in FASTA format, an assembly graph in GFA format, and a log file with detailed information about each stage. The GFA file is particularly useful for understanding why the assembly has certain structural features and for troubleshooting problematic regions.

Quality Assessment of Hybrid Assemblies

Genome Completeness

Genome completeness measures what fraction of expected genes are present in your assembly. The standard tool for this assessment is BUSCO, which compares your assembly against lineage-specific sets of single-copy orthologs. For bacteria, you should use the bacteria or your specific taxonomic group's BUSCO set.

A study comparing assembly tools for Mycobacterium tuberculosis genomes found that Unicycler-based assemblies had significantly higher genome completeness at approximately 98.7 percent compared to other assembler tools. This level of completeness is typical for high-quality bacterial assemblies, and values below 95 percent should trigger investigation into whether your input data or assembly parameters need adjustment.

Assembly Size and Contiguity

Compare your assembly size to the expected genome size for your species. The Mycobacterium tuberculosis study found that assembly sizes varied by assembler, with RagOut producing significantly longer assemblies at 4,418,574 base pairs compared to Unicycler at 4,377,642 base pairs and RagTag at 4,380,711 base pairs. The reference H37Rv genome is 4,411,532 base pairs.

Assembly size differences can indicate genuine biological variation or assembly errors. If your assembly is substantially larger than expected, check for contamination or misassembly. If it is substantially smaller, you may have collapsed repeats or missing regions.

Base Accuracy

Base accuracy in the final assembly depends on both the short-read accuracy and the depth of coverage. You can assess accuracy by mapping your short reads back to the assembly and examining mismatches. For finished genomes, you should aim for accuracy better than Q40, meaning fewer than one error per 10,000 bases.

The hybrid approach achieves high accuracy because the short reads provide the consensus sequence, while the long reads resolve the structure. This is why Unicycler can produce accurate assemblies even when long-read accuracy is low.

Plasmid Reconstruction

Plasmids are often the most challenging part of bacterial genome assembly. The MOB-suite was developed for reconstruction and typing of plasmids from draft assembly data. In testing against closed genomes with publicly available Illumina data, the MOB-suite identified contigs of plasmid origin with high sensitivity and specificity at 95 and 88 percent respectively.

For hybrid assemblies, you should expect better plasmid reconstruction than from short reads alone. The MOB-suite accurately reconstructed 207 of 377 known plasmids into single groupings, whereas plasmidSPAdes only reconstructed 102. However, the MOB-suite produced more hybrid plasmids, with 84 plasmids undergoing both splits and merges, indicating that plasmid reconstruction remains imperfect even with good assembly tools.

Comparing Assembly Tools

Unicycler versus Other Hybrid Assemblers

The Mycobacterium tuberculosis study compared Unicycler with RagOut and RagTag for hybrid assembly. Unicycler produced assemblies with the highest genome completeness, while RagOut produced assemblies with the fewest contigs and the longest genome size. The choice of assembler depends on your priorities.

If you prioritize completeness and accuracy, Unicycler is a strong choice. If you prioritize contiguity and are willing to accept potentially inflated genome sizes, RagOut may be preferable. For most applications, the completeness advantage of Unicycler outweighs the contiguity advantage of RagOut, because missing genes have more serious consequences than a slightly fragmented assembly.

SKESA for Short-Read-Only Assembly

SKESA is a DeBruijn graph-based de-novo assembler designed for assembling microbial genomes sequenced using Illumina. It produces assemblies with high sequence quality and contiguity, handles low-level contamination, and is fast. SKESA produces identical assemblies for the same input regardless of compute resources, which is valuable for reproducibility.

SKESA has been used for assembling over 272,000 read sets in the Sequence Read Archive at NCBI and for real-time pathogen detection. This scale of use demonstrates its reliability and suitability for surveillance applications. You might use SKESA for initial short-read-only assemblies and then use Unicycler when you have long-read data available.

When to Use Which Tool

For routine surveillance where you only have Illumina data, SKESA or SPAdes are appropriate. For producing finished reference genomes, hybrid assembly with Unicycler is the standard approach. For plasmid-focused analysis, you should use the MOB-suite in addition to your assembler of choice.

The choice also depends on your computational resources. SKESA is fast and produces identical results regardless of compute resources, which makes it suitable for high-throughput environments. Unicycler is more computationally intensive but produces more complete assemblies when long reads are available.

Troubleshooting Common Assembly Problems

Low Genome Completeness

If BUSCO completeness is below 95 percent, check your input data quality first. Low coverage, high error rates, or contamination can all reduce completeness. Re-examine your read trimming parameters and consider whether you need additional sequencing.

If your input data looks good, try adjusting Unicycler parameters. Increasing the k-mer sizes for the SPAdes step can help with some genomes, while decreasing them helps with others. You can also try assembling with different long-read depth by subsampling your long reads.

Fragmented Assemblies

Fragmentation indicates that the assembler could not resolve repetitive regions. This can happen when long-read coverage is too low or when the repeats are longer than the long reads. Check your long-read length distribution and consider whether you need additional sequencing with longer reads.

You can also examine the assembly graph to identify unresolved regions. The GFA file from Unicycler shows the graph structure, and you can identify bridges that connect different parts of the assembly. These bridges often represent repeats that need additional long-read coverage to resolve.

Inflated Genome Size

An assembly that is substantially larger than the expected genome size often indicates contamination. Check your input reads for contamination from other organisms, and examine the assembly for contigs with unusual coverage or GC content compared to the rest of the genome.

The MOB-suite can help identify plasmid contigs, which may legitimately increase assembly size. However, if you have contigs that do not match your expected species, you should investigate whether they represent contamination or genuine biological features such as prophages.

Chimeric Assemblies

Chimeric assemblies occur when the assembler incorrectly joins sequences from different parts of the genome or from different organisms. These are difficult to detect from assembly statistics alone. You can check for chimeras by examining read coverage across the assembly and by looking for regions where coverage drops dramatically.

Mapping your short reads back to the assembly and examining the alignment can reveal chimeric junctions. Reads should map consistently across the assembly without large gaps or orientation changes. If you find suspicious regions, you can manually inspect the assembly graph to understand the structure.

Records and Documentation for Assembly Projects

What to Record

For reproducible assembly projects, record the version of every tool you use, the parameters you select, and the input data characteristics. This includes the sequencing platform, read length distributions, coverage estimates, and quality metrics for both short and long reads.

Record the assembly statistics for each assembly attempt, including N50, total assembly size, number of contigs, and BUSCO completeness. This documentation allows you to compare different parameter sets and to understand why certain assemblies succeed while others fail.

Version Control and Reproducibility

Use version control for your assembly scripts and record the exact commands you run. SKESA produces identical assemblies for the same input regardless of compute resources, which makes it particularly suitable for reproducible workflows. Unicycler may produce slightly different results depending on the version and the underlying SPAdes version, so record these details.

Containerization with Docker or Singularity can improve reproducibility by fixing tool versions. This is especially important for long-term projects where tool versions may change and affect assembly results.

Data Management

Store raw sequencing data in a secure location with appropriate backup. For pathogen genomics, consider depositing raw reads in public repositories such as the Sequence Read Archive, which supports real-time pathogen detection through NCBI Pathogen Detection. This database integrates raw reads, assemblies, and metadata for public health surveillance.

Document the relationship between biological samples and sequencing data carefully. Sample mix-ups are a common source of errors in genomics projects, and clear labeling and tracking prevent these problems.

Biosafety and Laboratory Context

Risk Assessment for Sequencing Work

Working with bacterial cultures for sequencing requires appropriate biosafety practices. The Biosafety in Microbiological and Biomedical Laboratories manual from CDC and NIH provides guidance for risk assessment, containment, and decontamination. Before starting sequencing work, conduct a risk assessment for the organisms you are handling and follow the appropriate biosafety level practices.

For pathogenic bacteria, additional precautions may be necessary. The World Health Organization addresses antimicrobial resistance as a global health threat, and genomic surveillance of resistant pathogens is an important component of the response. If you are sequencing clinical isolates, follow your institution's protocols for handling potentially infectious materials.

Quality Control in the Laboratory

The American Society for Microbiology provides professional education and laboratory practice resources for microbiology. Following established laboratory practices for culture purity, identification, and storage ensures that your sequencing input is authentic and uncontaminated.

Use authenticated biological materials from culture collections such as the American Type Culture Collection when you need reference strains. Authenticated strains have documented provenance and quality testing, which is essential for method validation and for comparing results across laboratories.

Antimicrobial Resistance Context

Genomic assembly is a key step in antimicrobial resistance surveillance. The World Health Organization defines antimicrobial resistance as a global health priority and emphasizes surveillance and stewardship. Assembled genomes allow researchers to detect resistance genes and mutations that confer resistance, which informs treatment decisions and public health responses.

The Clinical and Laboratory Standards Institute provides standards for antimicrobial susceptibility testing, which remains the phenotypic reference for resistance determination. Genomic predictions of resistance should be validated against phenotypic testing when possible, because genomic methods can miss resistance mechanisms or predict resistance that is not phenotypically expressed.

Common Failure Patterns and How to Avoid Them

Failure Pattern 1: Insufficient Long-Read Coverage

The most common cause of poor hybrid assemblies is insufficient long-read coverage. When long-read coverage is too low, the assembler cannot resolve repetitive regions, and the assembly remains fragmented. This is particularly problematic for genomes with large repeats or multiple plasmids.

To avoid this failure, estimate your genome's repeat content before deciding on long-read sequencing depth. Genomes with many repeats need more coverage. You can also perform a test assembly with a subset of your long reads to assess whether you have enough data before committing to the full assembly.

Failure Pattern 2: Poor-Quality Input Reads

Assembly quality depends directly on input read quality. Low-quality short reads introduce base errors into the consensus sequence, while low-quality long reads may not align correctly to the assembly graph. Both problems degrade the final assembly.

Implement rigorous quality control before assembly. Trim adapters, filter low-quality reads, and check for contamination. If your reads fail quality metrics, resequence instead of proceeding with poor input data.

Failure Pattern 3: Ignoring Assembly Graph Structure

The assembly graph contains valuable information about the structure of your genome and the reasons for assembly difficulties. Ignoring the graph and relying only on FASTA output can hide problems that are visible in the graph structure.

Examine the GFA file from Unicycler to understand the assembly. Look for bridges, dead ends, and complex regions that may indicate unresolved repeats or misassemblies. This examination helps you decide whether additional sequencing or parameter adjustments are needed.

Failure Pattern 4: Overlooking Contamination

Contamination from other organisms or from the sequencing facility can produce assemblies with inflated genome sizes and spurious contigs. This is especially problematic for clinical or environmental samples where multiple organisms may be present.

Check for contamination early in your workflow. Examine k-mer distributions, map reads to your expected species, and inspect assembly statistics for anomalies. SKESA handles low-level contamination in reads, but you should still verify that your assembly represents the intended organism.

Professional Escalation Criteria

When to Seek Additional Expertise

If your assembly consistently fails quality metrics despite adjusting parameters and improving input data quality, seek assistance from colleagues with more experience in genome assembly. Assembly problems can be subtle, and an experienced eye can often identify issues that are not obvious from standard metrics.

For clinical or public health applications, consult with bioinformatics specialists who have experience with pathogen genomics. The NCBI Pathogen Detection system provides resources for pathogen genomics and can help you understand how your assemblies compare to others in the database.

When to Consider Alternative Approaches

If hybrid assembly with Unicycler does not produce satisfactory results, consider alternative approaches. Long-read-only assembly with tools designed for that purpose may work better for some genomes. Alternatively, optical mapping or other genome-mapping technologies can help resolve complex structural features.

For plasmid-focused analysis, the MOB-suite provides specialized tools for plasmid reconstruction and typing. Using these tools in addition to your assembler can improve plasmid characterization even when the overall assembly is imperfect.

When to Resequence

If your input data is fundamentally inadequate, resequencing is often the best option. This is preferable to spending extensive time trying to improve an assembly from poor-quality data. Consider resequencing when coverage is too low, read quality is poor, or contamination cannot be removed.

For clinical or public health applications, the cost of resequencing is usually justified by the importance of accurate results. A high-quality assembly from good data is more valuable than a questionable assembly from inadequate data.

A Practical Decision Framework for Hybrid Assembly Strategy Selection

Defining the Decision Problem

Bioinformaticians often treat hybrid assembly as a single workflow, but the choice of whether to use hybrid assembly at all, and how to allocate sequencing resources, depends on the biological question, the organism, and the downstream application. A structured decision framework helps you avoid wasting sequencing budget on unnecessary long-read data or producing assemblies that do not meet the quality requirements for your specific application.

The decision framework presented here is organized around three sequential questions. First, what is the minimum assembly quality required for your downstream analysis? Second, what are the genomic characteristics of your organism that affect assembly difficulty? Third, what sequencing resources are available and what is the cost trade-off between short-read depth, long-read depth, and alternative approaches?

Question 1: Define the Required Assembly Quality

Different downstream applications require different levels of assembly quality. You should define your quality threshold before sequencing, not after assembly, because this decision determines how much long-read data you need to generate.

For routine surveillance and screening applications, a high-quality draft assembly from short reads alone may be sufficient. The SKESA assembler was designed for assembling microbial genomes sequenced using Illumina and has been used for over 272,000 read sets in the Sequence Read Archive at NCBI and for real-time pathogen detection. If your application is detection of known resistance genes, MLST typing, or SNP-based clustering for outbreak investigation, a short-read-only assembly with high base accuracy is often adequate.

For reference genome production, comparative genomics, or structural variant analysis, you need a finished or near-finished assembly. The Mycobacterium tuberculosis study demonstrated that hybrid assemblies with ONT and PacBio long reads produced a greater number of detected genes, with a mean of 4,620 genes, compared to 4,478 genes from short-read assembly alone. If your downstream analysis depends on complete gene content, such as pangenome analysis or detection of large structural variants, hybrid assembly is necessary.

For plasmid-focused studies, the decision is more nuanced. The MOB-suite developers found that analyzing plasmid content from short-read draft assemblies remains difficult due to incomplete assembly of plasmids. If your research question concerns plasmid biology, transmission, or resistance gene localization, you should plan for hybrid assembly from the start.

Question 2: Assess Genomic Characteristics That Affect Assembly Difficulty

Before deciding on a sequencing strategy, assess the genomic features of your organism that influence assembly difficulty. These features determine how much long-read coverage you need and whether hybrid assembly will succeed.

The first characteristic is genome size. Bacterial genomes range from approximately 0.5 to 15 megabases, and larger genomes require more sequencing coverage for the same depth. The Mycobacterium tuberculosis genome is 4.4 megabases, which is typical for bacteria. For larger genomes, you need proportionally more reads to achieve the same coverage.

The second characteristic is repeat content. Genomes with multiple ribosomal RNA operons, insertion sequences, or prophages are more difficult to assemble because short reads cannot span these repeats. The Unicycler publication noted that the tool was designed to resolve repetitive regions using long-read information, but the amount of long-read coverage needed scales with repeat content. You can estimate repeat content from a preliminary short-read assembly by examining the assembly graph complexity or by running tools that identify repetitive regions.

The third characteristic is plasmid content. Bacteria can contain any number of plasmids, and assembly is complicated by the presence of repetitive elements in plasmids. The MOB-suite study found that bacterial isolates can contain any number of plasmids and that assembly remains complicated due to repetitive elements. If your organism is known to carry multiple plasmids, you should expect that plasmid reconstruction will require careful analysis and possibly additional sequencing.

The fourth characteristic is GC content and sequencing bias. High-GC organisms such as Mycobacterium tuberculosis can present challenges for short-read sequencing due to PCR bias. The Mycobacterium tuberculosis study used Illumina HiSeq, Oxford Nanopore PromethION, and PacBio platforms, and the choice of platform affected assembly outcomes. If your organism has extreme GC content, you may need to adjust sequencing protocols or use specific assembly parameters.

Question 3: Evaluate Sequencing Resources and Cost Trade-Offs

The third question concerns your available sequencing resources and the cost trade-offs between different strategies. This decision is often constrained by budget, instrument availability, and turnaround time.

Short-read sequencing with Illumina is the most cost-effective approach per base and provides high accuracy. For a typical bacterial genome, you can achieve 100-fold coverage for a relatively modest cost. The SKESA assembler is fast and produces high-quality assemblies from Illumina data, making it suitable for high-throughput environments.

Long-read sequencing with Oxford Nanopore or PacBio is more expensive per base and has higher error rates. However, the Unicycler publication showed that hybrid assembly can produce accurate, complete assemblies even when long-read depth and accuracy are low. This means you do not need to generate the same depth of long-read data as you would for a long-read-only assembly.

The cost trade-off depends on your specific situation. If you already have Illumina infrastructure and access to a Nanopore instrument, the marginal cost of adding modest long-read coverage may be low. If you need to send samples to a sequencing facility, the cost of a second sequencing run may be substantial, and you should weigh this against the value of a finished assembly.

A practical approach is to sequence a small amount of long-read data initially, assess whether it is sufficient to resolve the assembly graph, and then decide whether additional long-read sequencing is needed. This iterative approach avoids overspending on long-read data that may not be necessary for your genome.

Implementing the Decision Framework

The decision framework can be implemented as a structured workflow with clear decision points. The following steps provide a practical implementation that you can adapt to your specific situation.

Step 1: Define your downstream analysis requirements and the minimum assembly quality needed. Write down the specific quality metrics that matter for your application, such as BUSCO completeness, N50, or plasmid reconstruction accuracy. This step should be completed before any sequencing is performed.

Step 2: Gather information about your organism's genome characteristics. If you have a reference genome for your species, examine its repeat content, plasmid content, and GC content. If no reference is available, you can estimate these features from a preliminary short-read assembly.

Step 3: Determine your sequencing budget and available instruments. Calculate the cost of short-read-only, long-read-only, and hybrid approaches for your expected genome size and coverage requirements.

Step 4: Apply the decision rules. If your downstream analysis requires complete gene content or plasmid reconstruction, choose hybrid assembly. If your analysis can tolerate a draft assembly, short-read-only assembly may be sufficient. If your genome has high repeat content or multiple plasmids, allocate more of your budget to long-read sequencing.

Step 5: Sequence a small amount of long-read data initially if you choose hybrid assembly. Run a preliminary assembly and examine the assembly graph to determine whether the long-read coverage is sufficient. If the graph has unresolved regions, sequence additional long-read data.

Step 6: Document your decisions and the rationale behind them. This documentation is valuable for reproducibility and for justifying sequencing expenditures to funding agencies or institutional review boards.

A Scoring System for Assembly Strategy Selection

To make the decision framework more concrete, you can use a simple scoring system that quantifies the factors influencing assembly strategy. This system helps you compare different options systematically instead of relying on intuition.

Score your downstream analysis requirements on a scale of 1 to 5, where 1 means a draft assembly is acceptable and 5 means a finished assembly is required. Applications such as routine surveillance and resistance gene screening typically score 1 to 2. Comparative genomics and pangenome analysis score 3 to 4. Reference genome production and structural variant analysis score 5.

Score your organism's assembly difficulty on a scale of 1 to 5, where 1 means a simple genome with few repeats and no plasmids and 5 means a complex genome with many repeats and multiple plasmids. You can estimate this score from the number of contigs in a preliminary short-read assembly, the presence of known repetitive elements, or the plasmid content of related strains.

Score your sequencing budget flexibility on a scale of 1 to 5, where 1 means a fixed budget that cannot accommodate additional sequencing and 5 means you have flexibility to add sequencing as needed.

Add the three scores together. A total of 3 to 6 suggests that short-read-only assembly is appropriate. A total of 7 to 11 suggests that hybrid assembly is warranted but that you should start with modest long-read coverage. A total of 12 to 15 suggests that you should prioritize long-read sequencing and plan for a finished assembly.

This scoring system is a heuristic, not a definitive rule. You should adjust it based on your specific circumstances and the characteristics of your organism. However, it provides a structured way to make sequencing decisions that are often made informally.

Comparison of Assembly Strategies Across Application Types

To illustrate the decision framework, consider how different application types map to assembly strategies.

For outbreak investigation and pathogen surveillance, the primary goal is rapid identification of strains and detection of known resistance determinants. The NCBI Pathogen Detection system integrates raw reads, assemblies, and metadata for public health surveillance, and it relies on short-read data from many isolates. For this application, short-read-only assembly with SKESA or SPAdes is appropriate because the focus is on SNP-level comparisons and detection of known genes, not on complete genome structure.

For antimicrobial resistance surveillance, the World Health Organization emphasizes the importance of genomic surveillance for tracking resistance mechanisms. Assembled genomes allow researchers to detect resistance genes and mutations, but the resolution required depends on the specific resistance mechanism. Point mutations in chromosomal genes can be detected from short-read assemblies, while resistance genes located on plasmids may require hybrid assembly for complete characterization.

For reference genome production, the goal is a complete, accurate genome that serves as a reference for future studies. The Mycobacterium tuberculosis study demonstrated that hybrid assemblies produce more complete gene content than short-read-only assemblies. For this application, you should invest in sufficient long-read coverage to produce a finished assembly.

For plasmid-focused studies, the MOB-suite provides specialized tools for plasmid reconstruction and typing. The study found that the MOB-suite accurately reconstructed 207 of 377 known plasmids into single groupings, whereas plasmidSPAdes only reconstructed 102. If your research focuses on plasmid biology, you should use hybrid assembly and specialized plasmid analysis tools.

Records and Metrics for Decision Documentation

To make your assembly strategy decisions transparent and reproducible, maintain a decision record that documents the following information for each project.

Record the biological question and the downstream analysis requirements. This includes the specific quality metrics that your analysis requires, such as minimum BUSCO completeness, maximum contig count, or plasmid reconstruction accuracy.

Record the organism characteristics that influenced your decision. This includes genome size, estimated repeat content, plasmid content, and GC content. Note the source of this information, such as a reference genome or a preliminary assembly.

Record the sequencing strategy you chose and the rationale. This includes the sequencing platforms, target coverage for short and long reads, and the expected cost. Note any constraints that influenced your decision, such as budget limitations or instrument availability.

Record the actual sequencing data generated, including read counts, read length distributions, and coverage estimates for both short and long reads. This information is essential for interpreting assembly results and for troubleshooting problems.

Record the assembly results and quality metrics for each assembly attempt. This includes N50, total assembly size, number of contigs, BUSCO completeness, and plasmid reconstruction results. Compare these metrics to your predefined quality thresholds and document any discrepancies.

Record any deviations from your planned strategy and the reasons for those deviations. For example, if you needed additional long-read sequencing to resolve the assembly graph, document why the initial coverage was insufficient.

Common Failure Patterns in Strategy Selection

Several common failure patterns arise from poor strategy selection instead of from technical assembly problems. Recognizing these patterns helps you avoid them in future projects.

The first failure pattern is underinvesting in long-read sequencing for complex genomes. If your organism has high repeat content or multiple plasmids, modest long-read coverage may be insufficient to resolve the assembly graph. The Unicycler publication showed that the tool can work with low long-read depth, but there is a lower limit below which assembly quality degrades. If you consistently produce fragmented assemblies for a particular organism, you may need to increase long-read coverage.

The second failure pattern is overinvesting in long-read sequencing for simple genomes. If your organism has few repeats and no plasmids, short-read-only assembly may produce a draft assembly that is sufficient for your downstream analysis. Spending additional budget on long-read sequencing that does not improve your results is wasteful.

The third failure pattern is failing to define quality thresholds before sequencing. If you do not know what assembly quality you need, you cannot determine how much sequencing to generate. This often leads to either insufficient data and poor assemblies or excessive data and wasted budget.

The fourth failure pattern is ignoring plasmid content in strategy selection. Plasmids are often the most challenging part of bacterial genome assembly, and the MOB-suite study demonstrated that plasmid reconstruction from short-read data is incomplete. If your research involves plasmids, you should plan for hybrid assembly and specialized plasmid analysis from the start.

Professional Escalation Criteria for Strategy Decisions

If you are uncertain about the appropriate assembly strategy for your organism or application, seek advice from colleagues with relevant experience. This is particularly important for clinical or public health applications where the consequences of poor assembly quality can be serious.

Consult with bioinformatics specialists who have experience with your organism or with similar applications. The American Society for Microbiology provides professional education and laboratory practice resources that can help you connect with experts and access training materials.

If you are working with pathogenic organisms, consult the Biosafety in Microbiological and Biomedical Laboratories manual from CDC and NIH for guidance on risk assessment and containment. Your sequencing strategy should account for the biosafety requirements of your organism.

If your assembly strategy decisions have public health implications, consider how your results will be used. The NCBI Pathogen Detection system integrates genomic data for public health surveillance, and the quality of your assemblies affects the reliability of these surveillance efforts.

When to Resequence Based on Strategy Review

If your assembly results do not meet your predefined quality thresholds, review your strategy decisions before resequencing. Determine whether the problem is insufficient data, inappropriate assembly parameters, or a mismatch between your strategy and your application requirements.

If the problem is insufficient long-read coverage, resequencing with additional long-read data may resolve the assembly graph. The Mycobacterium tuberculosis study compared assemblies from Illumina, ONT, and PacBio data and found that the choice of platform affected assembly outcomes. If your initial long-read data came from one platform, consider whether the other platform might produce better results for your organism.

If the problem is a mismatch between your strategy and your application, you may need to change your approach instead of generate more data. For example, if you chose short-read-only assembly for a plasmid-focused study and cannot reconstruct plasmids, you may need to switch to hybrid assembly instead of simply generating more short-read data.

If the problem is poor input data quality, resequencing is often the best option. The SKESA assembler handles low-level contamination in reads, but severely contaminated or degraded samples may not produce useful assemblies regardless of the assembler or strategy. In these cases, the cost of resequencing is justified by the value of accurate results.

Frequently Asked Questions

What is the minimum long-read coverage needed for hybrid assembly with Unicycler?

Unicycler can produce useful hybrid assemblies even when long-read depth and accuracy are low, according to the original tool publication. The optimal coverage depends on your genome's repeat structure, but starting with 20 to 50-fold long-read coverage is reasonable. You can assess whether you have enough coverage by examining the assembly graph for unresolved regions.

How does Unicycler combine short and long reads?

Unicycler first builds an assembly graph from short reads using SPAdes, then uses a semi-global aligner to align long reads to this graph. The long-read alignments resolve ambiguous paths through the graph, allowing the assembler to produce larger contigs with fewer misassemblies than short-read-only approaches.

What is the difference between hybrid assembly and long-read-only assembly?

Hybrid assembly combines short reads for accuracy and long reads for structural resolution. Long-read-only assembly relies solely on long reads, which requires higher coverage to overcome platform error rates. Hybrid assembly is generally more cost-effective because you need less long-read data.

How do I assess the quality of a hybrid bacterial genome assembly?

Assess genome completeness with BUSCO, contiguity with N50 and contig counts, base accuracy by mapping reads back to the assembly, and plasmid reconstruction with tools like the MOB-suite. Compare assembly size to the expected genome size for your species and investigate any substantial discrepancies.

Why does my assembly contain more contigs than expected?

Excess contigs usually indicate unresolved repetitive regions or insufficient long-read coverage. Examine the assembly graph to identify the unresolved regions and consider whether additional long-read sequencing would help. Some genomes legitimately contain many plasmids or mobile elements that increase contig counts.

Can I use Unicycler for non-bacterial genomes?

Unicycler was designed for bacterial genomes, and its performance on other organisms is not well characterized. For eukaryotic genomes or other complex organisms, consider assemblers designed for those genome types. The semi-global alignment approach may not scale well to very large or complex genomes.

How does the MOB-suite help with plasmid analysis?

The MOB-suite provides modular tools for reconstruction and typing of plasmids from draft assembly data. It identifies contigs of plasmid origin, reconstructs plasmids into groupings, and provides replicon typing, relaxase typing, and prediction of conjugation potential. This is valuable because plasmid analysis from draft assemblies is difficult due to repetitive elements.

What should I do if my assembly fails quality checks?

First, verify that your input reads are high quality and free of contamination. Second, try adjusting assembly parameters such as k-mer sizes and coverage thresholds. Third, examine the assembly graph to understand structural problems. If these steps do not resolve the issues, consider resequencing or consulting with experienced bioinformaticians.

Related Bacteriology Guides

• Hybrid Genome Assembly: Combining Short and Long Reads for Better Results
• Which Bacteria Are Common to Raw Poultry? A Safety and Pathogen Guide
• Salmonellosis in Poultry: From Raw Chicken to Human Health Risks
• E. coli and Salmonella on Raw Chicken: Comparative Pathogenesis and Food Safety
• Salmonella in Poultry and Raw Chicken: Public Health Risks and Prevention

References and Further Reading

• Biosafety in Microbiological and Biomedical Laboratories. CDC and NIH.
• Microbiology Resources. American Society for Microbiology.
• Antimicrobial Resistance. World Health Organization.
• Bacteriological Analytical Manual. U.S. Food and Drug Administration.
• Antimicrobial Susceptibility Testing. Clinical and Laboratory Standards Institute.
• Culture Collections. American Type Culture Collection.
• NCBI Pathogen Detection. National Center for Biotechnology Information.
• SKESA: strategic k-mer extension for scrupulous assemblies.. Genome biology, 2018.
• MOB-suite: software tools for clustering, reconstruction and typing of plasmids from draft assemblies.. Microbial genomics, 2018.
• Comparisons of genome assembly tools for characterization of Mycobacterium tuberculosis genomes using hybrid sequencing technologies.. PeerJ, 2024.
• Unicycler: Resolving bacterial genome assemblies from short and long sequencing reads.. PLoS computational biology, 2017.

This article is educational and does not replace validated laboratory procedures, clinical diagnosis, institutional biosafety review, or public-health guidance.