How Ultra-Long Reads from Oxford Nanopore Transform Genome Assembly: From Gaps to Complete Chromosomes

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

Key Takeaways

Ultra-long reads, exceeding 100 kilobases and potentially reaching megabases, are critical for resolving repetitive genomic regions that confound short-read assemblers. These repeats, such as ribosomal RNA gene clusters and transposable elements, cause fragmentation and gaps in draft assemblies by preventing unambiguous read placement.
Achieving ultra-long reads necessitates meticulous high molecular weight DNA extraction protocols that minimize mechanical shearing, alongside library preparation methods like transposase-mediated fragmentation or optimized protocols that avoid fragmentation entirely. The quality of the starting DNA directly dictates the maximum achievable read length.
Assembly workflows benefit significantly from ultra-long reads, enabling chromosome-level contiguity and the resolution of complex structural variations. Adding even modest coverage (e.g., 5-fold) of ultra-long reads can more than double assembly contiguity metrics like NG50, as demonstrated by improvements from ~3 Mb to ~6.4 Mb in human genome assemblies.
For high-accuracy assemblies, a polishing step using complementary short-read sequencing data is essential to correct base-level errors inherent in long-read technologies, particularly in homopolymer regions. This combined approach can achieve assembly accuracy exceeding 99.8%.
The decision to employ ultra-long reads should be based on specific assembly goals, such as achieving telomere-to-telomere completeness or resolving large structural variants, rather than being a universal requirement for all genome sequencing projects. Bacterial genomes or projects focused on gene discovery may not necessitate this advanced approach.

Genome assembly quality is limited by the ability of sequencing reads to span repetitive regions. Short-read platforms produce fragments that are too brief to bridge long repeats, leaving assembly graphs fragmented and genomes incomplete. Oxford Nanopore Technologies can generate reads exceeding 100 kilobases, and with optimized protocols, reads of several megabases have been reported. These ultra-long reads allow assemblers to traverse repetitive elements, resolve structural variation, and produce chromosome-level assemblies that approach telomere-to-telomere completeness. This article explains when ultra-long reads are necessary, how to generate them in the laboratory, and how to integrate them into assembly workflows for maximum contiguity.

The Core Problem: Why Assemblers Leave Gaps

Repetitive Regions Defeat Short-Read Assembly

Genomes contain repeated sequences that appear identical across multiple locations. When sequencing reads are shorter than the repeat unit or the distance between adjacent repeats, the assembler cannot determine which copy of the repeat a read originated from. This ambiguity creates branching points in the assembly graph, forcing the assembler to stop and leave a gap. Ribosomal RNA gene clusters, telomeres, centromeres, transposable elements, and segmental duplications all present this challenge.

Short-read platforms typically produce reads of 150 to 300 base pairs. Most repeats in eukaryotic genomes are substantially longer than this, so short reads cannot span them. Even when paired-end information is available, the insert size of the library limits the distance that can be bridged. The result is a draft assembly with thousands of contigs and scaffolds, with the true chromosome structure remaining unknown.

Long Reads Provide the Bridge

Long-read sequencing platforms produce reads that are orders of magnitude longer than short reads. A read of 100 kilobases can span an entire repeat unit plus flanking unique sequence, allowing the assembler to place the repeat unambiguously. The longer the read, the more repeats it can bridge and the more contiguous the resulting assembly becomes.

The relationship between read length and assembly contiguity is direct. When researchers added approximately 5-fold coverage of ultra-long reads to a human genome assembly, the contiguity more than doubled compared to using standard long reads alone. The NG50 increased from roughly 3 megabases to approximately 6.4 megabases. This improvement came from the ability of ultra-long reads to span regions that standard long reads could not resolve.

What Counts as Ultra-Long

The term ultra-long refers to reads that exceed the typical length distribution of standard nanopore runs. Standard nanopore sequencing often produces N50 read lengths of 20 to 50 kilobases. Ultra-long protocols aim for N50 values above 100 kilobases, with individual reads reaching hundreds of kilobases or more.

One published protocol describes achieving N50 read lengths of 50 to 70 kilobases with mechanical shearing and 90 to 100 kilobases with transposase-mediated fragmentation. Another study optimized DNA extraction and library preparation to achieve average N50 read lengths of 80.57 kilobases, with maximum reads reaching 5.83 megabases. These lengths are sufficient to span most repetitive elements and produce near-complete chromosome assemblies.

At a Glance: Ultra-Long Read Assembly Decision Table

Scenario	Recommended Approach	Expected Outcome	Key Consideration
Bacterial genome under 10 Mb	Standard nanopore reads with 50 to 100x coverage	Single circular contig	Ultra-long reads optional, short reads can polish accuracy
Plant or animal genome with known complex repeats	Ultra-long reads with 30 to 60x coverage plus short-read polishing	Chromosome-scale scaffolds with few gaps	DNA extraction quality determines maximum read length
Telomere-to-telomere complete assembly	Ultra-long reads with adaptive sampling for gap closure	Complete chromosomes including telomeres and centromeres	Requires multiple library preparations and iterative gap filling
Structural variant detection in clinical samples	Ultra-long reads at 20 to 30x coverage	Breakpoint-level resolution of large variants	Sample input quantity may limit read length

Generating Ultra-Long Reads in the Laboratory

High Molecular Weight DNA Extraction Is the Foundation

The maximum read length achievable on a nanopore device is limited by the length of the DNA molecules loaded into the flow cell. If the DNA is fragmented during extraction, no amount of sequencing time will produce long reads. High molecular weight DNA extraction is therefore the first and most critical step.

Standard DNA extraction kits often shear DNA through pipetting, vortexing, or column-based purification. These methods are unsuitable for ultra-long read sequencing. Instead, protocols use gentle lysis, wide-bore pipette tips, and minimal handling to preserve DNA integrity. Some protocols recommend starting with 20 to 25 micrograms of high molecular weight DNA to achieve optimal read lengths.

Fresh or frozen cells can both serve as starting material. The published ultra-long read protocol describes extraction from both fresh and frozen mammalian cells, with careful attention to avoiding mechanical stress throughout the procedure. The quality of the starting material directly influences the maximum read length, so samples that have been repeatedly frozen and thawed or stored improperly will yield shorter reads.

Library Preparation Choices Affect Read Length

After extraction, the DNA must be prepared for sequencing. Two main approaches exist: mechanical shearing and transposase-based fragmentation. The choice affects both read length distribution and throughput.

Mechanical shearing produces a controlled size distribution but introduces a length ceiling. The published protocol reports N50 read lengths of 50 to 70 kilobases with mechanical shearing. This approach is appropriate when consistent fragment sizes are needed and when the maximum read length is less critical.

Transposase-mediated fragmentation achieves longer reads. The same protocol reports N50 read lengths of 90 to 100 kilobases with this method. Transposase-based approaches cleave DNA at specific sequence motifs, producing fragments that are then ligated to adapters. The reduced mechanical handling preserves longer molecules.

For maximum read length, some protocols minimize fragmentation entirely. The plant genome study that achieved N50 read lengths above 80 kilobases and maximum reads of 5.83 megabases used optimized DNA extraction and library preparation designed to avoid fragmentation at every step. This approach requires more hands-on time but produces the longest reads.

Sequencing Conditions That Preserve Long Reads

Once libraries are prepared, the sequencing run itself must be optimized. Nanopore devices sequence individual molecules as they pass through protein pores. Longer molecules take more time to traverse the pore, so run duration must be sufficient to capture them.

The pore size and voltage settings affect whether long molecules can pass through intact. Some protocols recommend specific flow cell types and sequencing kits that are optimized for ultra-long reads. The published human genome study used a custom protocol to generate reads with N50 above 100 kilobases and individual reads up to 882 kilobases.

Real-time monitoring during the run allows researchers to assess read length distribution as data accumulate. If the N50 is lower than expected, the run can be stopped and the library re-prepared or the flow cell reloaded. This flexibility is a practical advantage of nanopore sequencing compared to platforms that require complete runs before data are available.

Assembly Workflows for Ultra-Long Reads

Choosing an Assembler

Multiple assemblers can handle nanopore reads, but they differ in their ability to exploit ultra-long reads. The choice of assembler affects both contiguity and accuracy. Researchers should select an assembler that explicitly supports ultra-long reads and that has been validated on similar genome types.

The human genome study used a pipeline that incorporated ultra-long reads into an existing assembly, demonstrating that ultra-long reads can be added to improve contiguity even when the initial assembly was built from standard long reads. This two-stage approach is practical when ultra-long reads are generated after an initial assembly has already been produced.

For plant genomes, the telomere-to-telomere study combined ultra-long sequencing with adaptive sampling. Adaptive sampling allows the sequencer to selectively enrich for reads from specific genomic regions, which is useful for filling gaps that remain after initial assembly. This approach was used to fill remaining gaps in an Arabidopsis genome assembly and to resolve an unknown telomeric region in watermelon.

Read Alignment and Mapping

Before assembly, reads are often aligned to a reference genome or to the initial assembly. Minimap2 is a general-purpose alignment program designed for long reads. It can map reads of approximately 1 kilobase or longer at error rates around 15 percent, which matches the error profile of nanopore sequencing. It also handles assembly contigs and closely related full chromosomes of hundreds of megabases.

Minimap2 uses split-read alignment and concave gap costs to handle long insertions and deletions. It is substantially faster than earlier long-read aligners, making it practical for the large datasets generated by ultra-long sequencing. The software is available as open source, and its documentation provides usage examples for different read types.

Alignment serves multiple purposes in an ultra-long read workflow. It can assess read length distribution, identify chimeric reads, estimate coverage, and detect structural variants. For genome assembly, alignment to a close reference can guide scaffolding and gap closure.

Incorporating Short-Read Polishing

Nanopore reads have higher error rates than short reads, particularly in homopolymer regions. While ultra-long reads provide contiguity, they may introduce base-level errors. The human genome study reported that assembly accuracy exceeded 99.8 percent after incorporating complementary short-read sequencing data. This polishing step is essential for producing high-quality final assemblies.

The polishing workflow typically involves aligning short reads to the long-read assembly and using the short-read information to correct base errors. This step is computationally intensive but necessary for applications that require high accuracy, such as variant calling or gene annotation.

For researchers who need to learn these workflows, several training resources are available. The Galaxy Training Network provides accessible tutorials for genome assembly and polishing. The European Bioinformatics Institute offers training on bioinformatics data resources and analysis methods. The Carpentries teaches foundational computing skills that are useful for managing large sequencing datasets.

Case Studies: From Gaps to Complete Chromosomes

Human Genome Assembly with Ultra-Long Reads

The 2018 human genome study demonstrated the transformative effect of ultra-long reads on assembly contiguity. The researchers generated 91.2 gigabases of sequence data from the GM12878 cell line, representing approximately 30-fold theoretical coverage. De novo assembly of nanopore reads alone produced a contiguous assembly with NG50 around 3 megabases.

Adding approximately 5-fold coverage of ultra-long reads more than doubled the contiguity, increasing NG50 to about 6.4 megabases. The final assembly covered 85.8 percent of the reference genome. Ultra-long reads enabled assembly and phasing of the 4-megabase major histocompatibility complex locus in its entirety, measurement of telomere repeat length, and closure of gaps in the GRCh38 reference assembly.

This study established that ultra-long reads are an incremental improvement and a qualitative change in what assembly can achieve. Regions that were previously impossible to assemble, such as the MHC locus with its complex repeat structure, became tractable.

Plant Telomere-to-Telomere Assembly

Plant genomes present additional challenges due to polyploidy, large genome sizes, and high repeat content. The 2024 plant genome study optimized DNA extraction and library preparation to achieve DNA lengths exceeding 485 kilobases. The average N50 read length was 80.57 kilobases, with reads reaching up to 440 kilobases and maximum reads of 5.83 megabases.

The study demonstrated that combining ultra-long sequencing with adaptive sampling can effectively fill gaps during assembly. This approach successfully filled remaining gaps in a near-complete Arabidopsis genome assembly and resolved the sequence of an unknown telomeric region in watermelon. These results show that complete telomere-to-telomere assemblies are feasible across diverse plant species.

The practical implication is that researchers working on plant genomes should invest in DNA extraction optimization before sequencing. The quality of the starting DNA determines whether ultra-long reads can be generated, and this directly affects whether complete chromosome assemblies are achievable.

Viral Genome Assembly and Integration Detection

Ultra-long reads are also valuable for smaller genomes, particularly when structural complexity is involved. A 2021 study used nanopore sequencing to assemble human papillomavirus genomes from cervical cancer tissue and a cell line. The assembled HPV35 genome was 7,894 base pairs, sharing 99.96 percent identity with Sanger sequencing results. The HPV16 genome from the CaSki cell line was 7,904 base pairs, sharing 99.99 percent identity with the reference.

The key advantage of nanopore sequencing in this context was the ability to detect viral integration. Long reads directly revealed chimeric cellular-viral sequences and concatemeric genomic sequences, leading to the discovery of 448 unique integration breakpoints in the CaSki cell line and 60 breakpoints in the cervical cancer sample. Short-read sequencing struggles with this analysis because integration breakpoints often occur in repetitive regions.

This case demonstrates that ultra-long reads are valuable for large genomes and for resolving structural complexity in small genomes. Researchers studying viral integration, structural variation, or repetitive elements should consider whether ultra-long reads could resolve questions that short reads cannot answer.

Practical Implementation Steps

Step 1: Assess Whether Ultra-Long Reads Are Necessary

Not every assembly project requires ultra-long reads. Bacterial genomes, small viral genomes, and genomes with low repeat content can often be assembled to completion with standard long reads. The decision should be based on the genome size, repeat content, and the biological questions being asked.

If the goal is a complete chromosome-level assembly with telomeres and centromeres, ultra-long reads are necessary. If the goal is a draft assembly for gene discovery or comparative genomics, standard long reads may suffice. Researchers should evaluate the repeat content of their target genome using available databases or by examining the assembly graph from an initial long-read run.

Step 2: Optimize DNA Extraction

The most common cause of short reads is poor DNA quality. Researchers should test their DNA extraction protocol before committing to a full sequencing run. The DNA should be assessed by pulsed-field gel electrophoresis or a similar method to confirm that molecules exceed 100 kilobases.

Starting material should be fresh or properly frozen. Repeated freeze-thaw cycles fragment DNA. The extraction protocol should minimize pipetting, vortexing, and other mechanical stress. Wide-bore pipette tips and gentle mixing are essential.

Step 3: Prepare Libraries with Ultra-Long Read Protocols

Library preparation should follow protocols specifically designed for ultra-long reads. Transposase-based fragmentation generally produces longer reads than mechanical shearing. The published protocol describes achieving N50 read lengths of 90 to 100 kilobases with transposase-mediated fragmentation.

The amount of input DNA matters. The published protocol recommends 20 to 25 micrograms of high molecular weight DNA. Insufficient input DNA can lead to adapter contamination and reduced sequencing output.

Step 4: Sequence with Appropriate Run Conditions

The sequencing run should be configured for the expected read lengths. Longer reads require longer run durations. Real-time monitoring allows researchers to assess read length distribution and stop the run if the N50 is too low.

Flow cell quality and pore count affect throughput. Researchers should check the number of active pores before starting the run and consider reloading the flow cell if pore count is low.

Step 5: Assemble and Polish

The assembly workflow should include an assembler that supports ultra-long reads, followed by short-read polishing. The human genome study demonstrated that adding ultra-long reads to an existing assembly can more than double contiguity, so researchers with existing assemblies can benefit from generating ultra-long reads and incorporating them.

Polishing with short reads is essential for base-level accuracy. The human genome study reported accuracy exceeding 99.8 percent after short-read polishing. This step should not be skipped for applications that require high accuracy.

Step 6: Evaluate Assembly Quality

Assembly quality should be assessed using multiple metrics. Contiguity metrics such as N50 and NG50 indicate how much of the genome is contained in the largest contigs. Completeness metrics, such as BUSCO scores, indicate whether expected genes are present. Alignment to a close reference can identify misassemblies and structural errors.

For telomere-to-telomere assemblies, the presence of telomeric repeats at chromosome ends and the absence of internal gaps are key quality indicators. The plant genome study used these criteria to confirm complete assemblies.

Records and Measurements

Read Length Statistics

The N50 read length is the primary metric for assessing whether ultra-long reads have been achieved. N50 is the length at which 50 percent of the total sequenced bases are in reads of that length or longer. An N50 above 100 kilobases indicates that the ultra-long read protocol is working.

Maximum read length is also informative, as it indicates the upper limit of what the DNA extraction and library preparation can achieve. The plant genome study reported maximum reads of 5.83 megabases, demonstrating that megabase-scale reads are possible with optimized protocols.

Coverage Calculations

Coverage is calculated as the total number of sequenced bases divided by the genome size. The human genome study used approximately 30-fold coverage of standard nanopore reads plus an additional 5-fold coverage of ultra-long reads. The viral genome study used much higher coverage, with 97-fold coverage for HPV35 and 3,857-fold coverage for HPV16.

For ultra-long read assembly, the distribution of coverage matters as much as the total. Ultra-long reads are more difficult to generate, so they may be present at lower coverage than standard reads. The human genome study showed that even 5-fold coverage of ultra-long reads can dramatically improve contiguity.

Assembly Metrics

NG50 is the contiguity metric that accounts for genome size. It is the length at which 50 percent of the genome is contained in contigs of that length or longer. The human genome study reported NG50 values of approximately 3 megabases with standard reads and 6.4 megabases with ultra-long reads added.

The number of contigs and the number of gaps are also important. A complete chromosome assembly should have one contig per chromosome, with no internal gaps. The plant genome study achieved this for Arabidopsis and watermelon, demonstrating that complete assemblies are achievable.

Common Failure Patterns

DNA Fragmentation During Extraction

The most common cause of failed ultra-long read runs is DNA fragmentation during extraction. Symptoms include N50 read lengths well below 50 kilobases and a maximum read length that is far below expectations. The solution is to revise the extraction protocol, reduce mechanical stress, and verify DNA integrity before sequencing.

Insufficient Input DNA

Ultra-long read protocols require substantial input DNA. The published protocol recommends 20 to 25 micrograms. If the input is insufficient, the library may have adapter contamination, low sequencing output, or poor read length distribution. Researchers should quantify DNA accurately and concentrate samples if needed.

Flow Cell Quality Issues

Nanopore flow cells have a limited number of pores, and pore quality varies between batches. A flow cell with few active pores will produce low throughput regardless of library quality. Researchers should check pore count before starting the run and consider using a new flow cell if pore count is low.

Chimeric Read Formation

Chimeric reads, where two DNA molecules are joined during library preparation, can confuse assemblers. These reads appear to span regions that are not actually adjacent in the genome. Alignment to a reference can identify chimeric reads, and assemblers may have filters to exclude them. The frequency of chimeric reads increases with library preparation errors, so careful protocol adherence is important.

Insufficient Coverage of Ultra-Long Reads

Ultra-long reads are more difficult to generate than standard reads, so they may be present at lower coverage. If ultra-long read coverage is too low, the assembler may not have enough data to resolve repeats. The human genome study showed that 5-fold coverage of ultra-long reads was sufficient to improve contiguity, but this may vary by genome.

Limitations and Interpretation

Error Rates in Homopolymer Regions

Nanopore sequencing has higher error rates than short-read platforms, particularly in homopolymer regions where the same nucleotide is repeated. These errors can affect base-level accuracy even when the assembly is contiguous. Short-read polishing is essential for correcting these errors, but it may not resolve all homopolymer issues.

Coverage Bias and Unevenness

Ultra-long reads are not uniformly distributed across the genome. Some regions may have very high coverage while others have very low coverage. This unevenness can affect assembly quality and variant calling. Researchers should examine coverage distribution and consider whether additional sequencing is needed for underrepresented regions.

Computational Requirements

Assembling ultra-long reads requires substantial computational resources. The alignment and assembly steps are memory-intensive, and polishing adds additional computational burden. Researchers should ensure that their computing infrastructure can handle the data volume before starting.

Reference Bias in Assessment

Assessing assembly quality by alignment to a reference genome can introduce bias. If the reference has errors or structural differences from the sample, the assessment may underestimate assembly quality. Researchers should use multiple assessment methods, including reference-free metrics such as BUSCO and k-mer analysis.

Safety and Regulatory Context

Data Management and Privacy

Genome sequencing data can contain sensitive information, particularly for human samples. Researchers must follow institutional and regulatory requirements for data storage, access, and sharing. The NCBI provides databases and tools for sequence data management, and researchers should be familiar with the data submission and access policies.

Sample Handling and Biosafety

DNA extraction from biological samples requires appropriate biosafety practices. Researchers should follow institutional guidelines for handling potentially infectious materials. The specific requirements depend on the sample type and the pathogens that may be present.

Reproducibility and Documentation

Reproducible research requires detailed documentation of protocols, parameters, and data processing steps. Workflow management systems such as nf-core provide standardized pipelines that improve reproducibility. The Bioconductor project offers packages for reproducible genomic analysis. Researchers should document their methods thoroughly to enable replication and verification.

Professional Escalation Criteria

When to Seek Expert Assistance

Researchers should consider consulting with bioinformatics experts or core facilities when they encounter persistent problems with ultra-long read generation or assembly. Specific situations that warrant escalation include:

N50 read lengths remain below 50 kilobases despite protocol optimization
Assembly contiguity does not improve when ultra-long reads are added
The assembler crashes or produces inconsistent results across runs
The assembly contains many misjoins or structural errors that cannot be resolved
The computational requirements exceed local infrastructure capacity

Training and Skill Development

Researchers who are new to ultra-long read sequencing should invest in training before starting their projects. The Galaxy Training Network provides tutorials for genome assembly and analysis. The European Bioinformatics Institute offers training on bioinformatics data resources. The Carpentries teaches foundational computing skills that are useful for managing sequencing data.

The nf-core documentation describes community standards for reproducible pipelines, which can help researchers adopt best practices. The Bioconductor project provides packages for genomic analysis with documented workflows.

A Practical Decision Framework for Matching Ultra-Long Read Effort to Assembly Goals

The Cost-Benefit Question Every Assembly Project Faces

Ultra-long read generation demands specialized DNA extraction, careful library preparation, and extended sequencing runs. These requirements translate into additional labor, consumable costs, and optimization cycles. Before committing resources, researchers need a structured way to determine whether ultra-long reads will solve their specific assembly problem or whether standard long reads will suffice. The decision framework below provides a systematic method for matching sequencing effort to assembly goals, based on genome characteristics and the biological questions at hand.

Step 1: Characterize the Target Genome Before Sequencing

The first decision point occurs before any sequencing begins. Researchers should compile available information about their target genome to predict whether ultra-long reads will be necessary. Key characteristics to assess include genome size, repeat content, ploidy level, and the presence of known complex regions such as ribosomal RNA clusters, centromeres, or telomeres.

For well-studied species, published genome assemblies and repeat databases provide a starting point. The NCBI hosts reference genomes and repeat annotations for thousands of species, allowing researchers to examine the repeat landscape of closely related organisms. If a related species has been assembled to chromosome level, the repeat structure of that assembly indicates what challenges the target genome may present.

For less-studied organisms, a preliminary low-coverage long-read run can provide empirical data. A single flow cell producing 10 to 20-fold coverage of standard long reads can reveal the repeat structure through the assembly graph. If the graph contains many unresolved branches or if the N50 of the resulting assembly is far below the expected chromosome size, ultra-long reads are likely needed.

Step 2: Classify the Assembly Goal

Assembly projects fall into distinct categories with different requirements for read length. The classification below helps researchers match their goals to appropriate sequencing strategies.

Draft assembly for gene discovery or comparative genomics. This goal requires a reasonably complete representation of coding regions and overall genome structure, but does not require complete chromosomes. Standard long reads with N50 values of 20 to 50 kilobases are usually sufficient. The assembly will contain gaps in repetitive regions, but these gaps rarely affect gene-level analyses.

Chromosome-scale assembly with scaffolds. This goal requires ordering and orienting contigs into chromosome-length scaffolds. Ultra-long reads can help bridge gaps between contigs, but alternative approaches such as Hi-C or optical mapping can also provide scaffolding information. If ultra-long reads are already being generated for other purposes, they can be incorporated into the scaffolding process.

Complete telomere-to-telomere assembly. This goal requires one contiguous sequence per chromosome, including telomeres, centromeres, and all repetitive regions. Ultra-long reads are essential for this goal, and even then, multiple rounds of sequencing and gap filling may be necessary. The plant genome study that achieved telomere-to-telomere assemblies used optimized DNA extraction to achieve N50 read lengths above 80 kilobases and combined ultra-long sequencing with adaptive sampling to fill remaining gaps.

Structural variant detection. This goal requires reads long enough to span the structural variants of interest. The required read length depends on the size of the variants. For variants larger than 50 kilobases, ultra-long reads provide a clear advantage. For smaller variants, standard long reads may suffice.

Step 3: Estimate the Repeat Complexity That Must Be Spanned

The core question for any assembly project is whether the reads can span the repetitive regions that cause assembly gaps. The required read length depends on the size of the repeats and the distance between adjacent copies.

For a repeat that is 10 kilobases long, a read of 20 kilobases can span the repeat plus 5 kilobases of flanking unique sequence on each side. This is sufficient to place the repeat unambiguously. For a repeat cluster spanning 100 kilobases, a read of 150 kilobases or longer is needed.

Researchers can estimate the repeat complexity of their target genome by examining repeat annotations from related species or by analyzing the assembly graph from a preliminary run. The key metric is the length of the longest repeat that must be spanned. If this length exceeds the N50 of standard long reads, ultra-long reads are necessary.

The human genome study provides a concrete example. Standard nanopore reads with an N50 around 20 to 30 kilobases produced an assembly with NG50 of approximately 3 megabases. Adding ultra-long reads with N50 above 100 kilobases more than doubled the contiguity to approximately 6.4 megabases. The improvement came from the ability of ultra-long reads to span repeats that were longer than the standard reads.

Step 4: Calculate the Required Ultra-Long Read Coverage

Once the decision to generate ultra-long reads is made, the next question is how much coverage is needed. The human genome study used approximately 5-fold coverage of ultra-long reads in addition to 30-fold coverage of standard reads. This modest amount of ultra-long read coverage was sufficient to more than double contiguity.

The optimal coverage depends on the repeat complexity and the assembly strategy. For genomes with moderate repeat content, 5 to 10-fold coverage of ultra-long reads may be sufficient. For genomes with extensive repeat arrays or large segmental duplications, higher coverage may be needed.

A practical approach is to generate ultra-long reads in batches and assess assembly improvement after each batch. The real-time nature of nanopore sequencing allows researchers to monitor read length distribution as data accumulate and to decide whether additional sequencing is needed. This iterative approach avoids the cost of generating excessive coverage that does not improve the assembly.

Step 5: Choose Between One-Pass and Two-Stage Assembly

Ultra-long reads can be incorporated into assembly in two ways. The one-pass approach uses ultra-long reads as the primary data for a de novo assembly. The two-stage approach generates an initial assembly from standard long reads, then adds ultra-long reads to bridge gaps and improve contiguity.

The human genome study used a two-stage approach. The initial assembly from standard nanopore reads had an NG50 of approximately 3 megabases. Adding 5-fold coverage of ultra-long reads increased the NG50 to approximately 6.4 megabases. This approach is practical when ultra-long reads are generated after an initial assembly has been produced, or when the ultra-long read protocol is still being optimized.

The one-pass approach is simpler but requires that ultra-long reads be generated before assembly begins. This approach is appropriate when the ultra-long read protocol is well established and the researcher is confident in achieving the required read lengths.

The choice between one-pass and two-stage assembly depends on the timeline and the confidence in the ultra-long read protocol. If the protocol is new or unoptimized, the two-stage approach allows the researcher to generate an initial assembly while optimizing the ultra-long read protocol. If the protocol is well established, the one-pass approach saves time and computational resources.

Step 6: Plan for Iterative Gap Filling

Even with ultra-long reads, complete telomere-to-telomere assembly often requires iterative gap filling. The plant genome study demonstrated that combining ultra-long sequencing with adaptive sampling can effectively fill gaps during assembly. Adaptive sampling allows the sequencer to selectively enrich for reads from specific genomic regions, which is useful for targeting regions that remain unassembled after initial sequencing.

The iterative process involves assembling with available reads, identifying remaining gaps, generating additional ultra-long reads targeted to those gaps, and reassembling. This process can be repeated until the assembly is complete or until diminishing returns make further sequencing impractical.

For researchers pursuing complete assemblies, the decision framework should include a plan for iterative gap filling from the outset. This plan should specify the criteria for deciding when a gap is resolved, the methods for identifying gap locations, and the sequencing strategy for generating targeted reads.

Step 7: Document Decisions and Outcomes

A record system for assembly decisions helps researchers track what was tried, what worked, and what did not. The following records should be maintained for each assembly project:

Genome characteristics. Record the estimated genome size, repeat content, and ploidy level. Note the source of this information, whether from published data or preliminary sequencing.

Read length statistics. Record the N50, maximum read length, and read length distribution for each sequencing run. This information is essential for assessing whether the ultra-long read protocol is working and for comparing runs.

Assembly metrics. Record the NG50, number of contigs, number of gaps, and BUSCO completeness for each assembly iteration. These metrics provide a quantitative basis for deciding whether additional sequencing is needed.

Protocol parameters. Record the DNA extraction method, library preparation method, input DNA amount, and sequencing conditions for each run. This information allows researchers to identify which protocol changes improved read length and which had no effect.

Cost and time tracking. Record the consumable costs, instrument time, and personnel time for each sequencing run and assembly iteration. This information helps researchers estimate the resources required for future projects.

The Galaxy Training Network provides tutorials on genome assembly that include guidance on documenting assembly workflows. The nf-core documentation describes standards for reproducible pipelines that can be adapted for record keeping.

Common Failure Patterns in the Decision Framework

Overestimating the need for ultra-long reads. Some researchers generate ultra-long reads when standard long reads would suffice. This mistake wastes time and resources. The decision framework helps avoid this by requiring an explicit assessment of repeat complexity before committing to ultra-long read generation.

Underestimating the difficulty of ultra-long read generation. Ultra-long read protocols require careful attention to DNA quality and library preparation. Researchers who underestimate this difficulty may generate reads that are not substantially longer than standard reads, defeating the purpose of the effort. The record system helps identify protocol problems early.

Generating insufficient ultra-long read coverage. The human genome study showed that 5-fold coverage of ultra-long reads was sufficient to improve contiguity, but this may not hold for all genomes. Researchers who generate too little coverage may see little improvement and conclude that ultra-long reads are not helpful, when the real problem is insufficient data.

Ignoring the need for short-read polishing. Ultra-long reads provide contiguity but may introduce base-level errors. The human genome study reported that assembly accuracy exceeded 99.8 percent after incorporating short-read polishing. Researchers who skip this step may produce assemblies with high contiguity but low accuracy.

Professional Escalation Criteria for the Decision Framework

Researchers should seek expert assistance when the decision framework leads to persistent problems that cannot be resolved through protocol optimization. Specific situations that warrant escalation include:

The repeat complexity assessment suggests ultra-long reads are needed, but the ultra-long read protocol consistently produces N50 values below 50 kilobases despite multiple optimization attempts
The assembly contiguity does not improve when ultra-long reads are added, even at coverage levels that should be sufficient based on the repeat complexity assessment
The iterative gap filling process does not converge, with the same gaps remaining after multiple rounds of targeted sequencing
The computational requirements for assembling ultra-long reads exceed local infrastructure capacity, requiring access to high-performance computing resources

The European Bioinformatics Institute offers training on bioinformatics data resources that can help researchers build the skills needed to troubleshoot assembly problems. The Bioconductor project provides packages for genomic analysis with documented workflows that can support assembly evaluation.

Integrating the Decision Framework with Existing Workflows

The decision framework is designed to complement, not replace, existing assembly workflows. Researchers should integrate the framework into their project planning from the outset, using it to guide decisions about sequencing strategy, resource allocation, and timeline.

The framework is particularly valuable for projects with limited resources, where the cost of unnecessary ultra-long read generation can be substantial. By providing a structured method for assessing whether ultra-long reads are needed, the framework helps researchers allocate resources to the steps that will have the greatest impact on assembly quality.

For researchers working in core facilities or service laboratories, the framework provides a basis for discussing project requirements with clients. The classification of assembly goals and the repeat complexity assessment give both the researcher and the client a common language for describing what the project requires and what outcomes can be expected.

The Carpentries teaches foundational computing skills that are useful for implementing the record system and for analyzing assembly metrics. These skills include shell scripting, data management, and version control, all of which support the reproducible documentation that the decision framework requires.

Frequently Asked Questions

What is the difference between long reads and ultra-long reads?

Long reads typically have N50 values of 10 to 50 kilobases, while ultra-long reads have N50 values above 100 kilobases. The distinction matters because ultra-long reads can span larger repetitive regions and produce more contiguous assemblies. The human genome study showed that adding ultra-long reads more than doubled assembly contiguity compared to standard long reads alone.

How much DNA is needed for ultra-long read sequencing?

The published ultra-long read protocol recommends 20 to 25 micrograms of high molecular weight DNA. This amount is higher than what is needed for standard nanopore sequencing because the library preparation steps for ultra-long reads have lower efficiency. Insufficient input DNA can lead to adapter contamination and reduced read length.

Can ultra-long reads be added to an existing assembly?

Yes. The human genome study demonstrated that adding approximately 5-fold coverage of ultra-long reads to an existing assembly more than doubled contiguity. This two-stage approach is practical when ultra-long reads are generated after an initial assembly has been produced. The ultra-long reads are used to bridge gaps and resolve regions that the initial assembly could not span.

What is adaptive sampling and how does it help with gap filling?

Adaptive sampling is a nanopore sequencing feature that allows the sequencer to selectively enrich for reads from specific genomic regions. The plant genome study used adaptive sampling to fill remaining gaps in an Arabidopsis genome assembly and to resolve an unknown telomeric region in watermelon. This approach is useful when specific regions remain unassembled after initial sequencing.

Why is short-read polishing necessary if ultra-long reads are so accurate?

Nanopore reads have higher error rates than short reads, particularly in homopolymer regions. While ultra-long reads provide contiguity, they may introduce base-level errors. The human genome study reported that assembly accuracy exceeded 99.8 percent after incorporating short-read polishing. This step is essential for applications that require high base-level accuracy.

How do I know if my DNA extraction is good enough for ultra-long reads?

The most reliable test is to assess DNA integrity by pulsed-field gel electrophoresis or a similar method. DNA molecules should exceed 100 kilobases for ultra-long read sequencing. If the DNA is fragmented, the N50 read length will be low regardless of sequencing conditions. Researchers should verify DNA integrity before investing in library preparation and sequencing.

What assemblers work best with ultra-long reads?

Multiple assemblers can handle nanopore reads, but they differ in their ability to exploit ultra-long reads. Researchers should select an assembler that explicitly supports ultra-long reads and that has been validated on similar genome types. The human genome study used a pipeline that incorporated ultra-long reads into an existing assembly, demonstrating that this approach can work with standard tools.

How much coverage of ultra-long reads is needed?

The human genome study used approximately 5-fold coverage of ultra-long reads in addition to 30-fold coverage of standard nanopore reads. This was sufficient to more than double assembly contiguity. The optimal coverage depends on the genome size, repeat content, and assembly goals. Researchers may need to experiment to find the right balance for their specific project.

Related Bioinformatics Guides

Related Clinical & Scientific Guides

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.
Nanopore sequencing and assembly of a human genome with ultra-long reads.. Nature biotechnology, 2018.
Minimap2: pairwise alignment for nucleotide sequences.. Bioinformatics (Oxford, England), 2018.
Whole Genome Assembly of Human Papillomavirus by Nanopore Long-Read Sequencing.. Frontiers in genetics, 2021.
Nanopore ultra-long sequencing and adaptive sampling spur plant complete telomere-to-telomere genome assembly.. Molecular plant, 2024.
Ultra-long Read Sequencing for Whole Genomic DNA Analysis.. Journal of visualized experiments : JoVE, 2019.

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