Quality Control Metrics for Long-Read Sequencing: What to Check Beyond Read Length and N50

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

Long-read sequencing platforms from Pacific Biosciences and Oxford Nanopore Technologies produce data that differs fundamentally from short-read Illumina output. Read length and N50 statistics describe throughput and contiguity potential, but they do not reveal whether individual reads are accurate enough for variant calling, whether adapter sequences contaminate the data, whether chimeric reads will corrupt assembly graphs, or whether base modification signals are preserved. This article defines the quality control metrics that matter for PacBio and Nanopore data, explains how to interpret them in the context of specific downstream applications, and provides practical thresholds and tool recommendations for laboratory professionals and bioinformatics analysts.

Key Takeaways

• Long-read sequencing quality control must assess four layers beyond read length and N50: raw signal quality, read-level accuracy, library-level contamination, and assembly or alignment-level consistency, each answering a different question about the data.
• Chimeric reads, which join sequence from two genomic locations and arise from library preparation artifacts or basecalling errors, are not reported by standard summary statistics and must be detected by alignment-based methods or assembly graph inspection.
• Adapter contamination in long-read data reduces effective coverage, introduces spurious alignments, and can create false structural variant calls, so screening for adapter sequences should be performed before assembly or alignment.
• Coverage requirements vary by platform, with PacBio HiFi genome assembly of yeast requiring more than 20-fold coverage while ONT data required more than 80-fold coverage for high-quality assemblies.
• Base modification signals from Oxford Nanopore and PacBio HiFi sequencing can influence basecalling accuracy and create systematic errors, so verifying their presence and distribution is critical for epigenetic studies and recommended for standard analyses.

The Scope of Long-Read Quality Control

Quality control for long-read sequencing extends beyond the summary statistics printed by the sequencer. A complete QC workflow must assess four layers of data quality: raw signal or basecall quality, read-level accuracy, library-level contamination, and assembly or alignment-level consistency. Each layer answers a different question about the data. Raw signal quality tells you whether the instrument performed correctly. Read-level accuracy tells you whether individual molecules were sequenced faithfully. Library-level contamination tells you whether adapters, barcodes, or foreign DNA compromise the dataset. Assembly or alignment-level consistency tells you whether the reads support the biological conclusions you intend to draw.

The choice of metrics depends on the sequencing platform. PacBio HiFi reads are circular consensus sequences with high per-base accuracy, while PacBio continuous long reads (CLR) are single-pass reads with lower accuracy. Oxford Nanopore reads have platform-specific error profiles that include homopolymer errors and base modification signals. A QC pipeline that works for one platform may miss critical failures in another. Tools such as LongReadSum provide cross-platform QC reports that cover read, base, base quality, alignment, and base modification metrics across PacBio, Oxford Nanopore, and Illumina Complete Long Read formats, which is useful when a laboratory runs multiple platforms or transitions between them [<a href="#ref-1">1</a>].

The practical outcome of rigorous QC is straightforward: you avoid wasting compute time and money on failed datasets, you detect contamination before it corrupts biological conclusions, and you select the correct assembly or variant calling strategy based on the actual quality of your reads. For clinical applications, QC metrics also determine whether a variant call is reportable or requires orthogonal confirmation [<a href="#ref-2">2</a>].

Core Quality Metrics for Long-Read Data

Read Length Distribution Beyond N50

N50 is the length at which 50 percent of the assembled or sequenced bases are contained in reads of that length or longer. It is a useful single number, but it hides the shape of the read length distribution. A dataset with a high N50 can still contain many short reads that complicate assembly or reduce mapping confidence. The full length distribution, including the mean, median, N90, and the fraction of reads above application-specific length thresholds, provides a more complete picture.

For structural variant detection, the fraction of reads that span the full length of a repetitive element or a large deletion matters more than the N50. For transcriptome analysis, the read length distribution relative to the expected transcript lengths determines whether full-length isoforms are captured. For genome assembly, the distribution of read lengths relative to repeat lengths determines whether the assembler can resolve repetitive regions. The Flye assembler, for example, constructs a repeat graph from error-prone long reads, and the quality of that graph depends on having enough reads that span each repeat junction [<a href="#ref-3">3</a>].

Per-Base Quality Scores and Their Limitations

Quality scores in long-read data are computed differently from short-read data. PacBio HiFi reads have quality scores that reflect the consensus accuracy of the circular consensus sequence. Oxford Nanopore basecallers produce per-read quality scores that estimate the probability of error at each position. These scores are useful for filtering, but they are not perfectly calibrated. A read with a high average quality score can still contain localized error clusters, particularly in homopolymer regions or at methylation sites.

The practical approach is to examine the distribution of quality scores across each read and across the dataset, beyond the mean. Tools that summarize base quality metrics across the entire dataset, such as LongReadSum, allow you to identify whether quality degrades toward read ends or in specific sequence contexts [<a href="#ref-1">1</a>]. For PacBio HiFi data, quality scores above Q20 (99 percent accuracy) are typical, but the distribution matters for variant calling in difficult regions.

Read Accuracy and Error Profiles

Read accuracy is the fraction of bases in a read that match the true template sequence. For PacBio HiFi reads, accuracy is high because each base is observed multiple times in the circular consensus. For Oxford Nanopore reads, accuracy depends on the basecaller version, the pore version, and the library preparation method. Error profiles differ by platform: Nanopore errors are dominated by insertions and deletions in homopolymer regions, while PacBio CLR errors are more evenly distributed between substitutions and indels.

The error profile determines which downstream tools will work well. Assemblers that model platform-specific error profiles, such as Flye for error-prone reads, can produce high-quality assemblies from reads that would fail with short-read assemblers [<a href="#ref-3">3</a>]. Polishing tools correct residual errors after assembly, and the choice of polisher depends on the platform. For yeast genomes assembled from ONT data, polishing with Pilon and Medaka improved accuracy and continuity respectively, and the combination of both worked well across most quality metrics [<a href="#ref-4">4</a>].

Chimerism Detection

Chimeric reads are reads that contain sequence from two different genomic locations joined together. They arise from library preparation artifacts, such as template switching during amplification, or from basecalling errors that join two molecules. Chimeric reads are particularly dangerous for genome assembly because they create false connections in the assembly graph. They are also problematic for structural variant calling because they mimic genuine genomic rearrangements.

Chimerism is not reported by standard sequencing summary statistics. You must detect it by aligning reads to a reference genome and looking for reads that map to two distant locations with a sharp transition, or by examining assembly graphs for unusual branching patterns. For de novo assembly projects, the repeat graph structure can reveal chimeric connections, but detecting them requires careful inspection of the assembly graph instead of a simple QC metric [<a href="#ref-3">3</a>].

Adapter Contamination

Adapter sequences are added during library preparation and must be removed before analysis. In long-read sequencing, adapter contamination can occur when the library preparation fails to remove adapters completely, when barcodes are not trimmed correctly, or when the sequencing run produces short reads that consist mostly of adapter sequence. Adapter contamination reduces the effective coverage, introduces spurious alignments, and can create false structural variant calls.

The standard approach is to screen reads for adapter sequences using alignment-based or k-mer-based methods. The fraction of reads containing adapter sequence and the fraction of bases that are adapter-derived should be reported. For barcoded libraries, you must also verify that barcode sequences are correctly assigned and that barcode hopping or cross-talk between samples is below acceptable thresholds.

Base Modification and Methylation Signals

Oxford Nanopore sequencing detects base modifications directly from the electrical current signal, and PacBio HiFi sequencing can detect methylation from the kinetics of the polymerase. These signals are biologically important for epigenetic studies, but they also affect QC. Base modification signals can influence basecalling accuracy, and the presence of modifications can create systematic errors in certain sequence contexts.

Tools that summarize base modification information across the dataset, such as LongReadSum, allow you to verify that modification signals are present and that they are distributed as expected [<a href="#ref-1">1</a>]. For methylation studies, the fraction of reads with detectable modification signals and the consistency of modification calls across reads are critical QC metrics. For standard genome assembly or variant calling, base modifications are usually ignored, but you should verify that they do not introduce systematic errors in your analysis.

Coverage Depth and Uniformity

Coverage depth is the average number of reads that cover each position in the genome. For long-read sequencing, coverage requirements depend on the application and the platform. For PacBio HiFi genome assembly of yeast, more than 20-fold coverage was sufficient for high-quality assemblies, while ONT data required more than 80-fold coverage [<a href="#ref-4">4</a>]. These thresholds are specific to the organism and the assembler, but they illustrate the general principle that coverage requirements vary by platform and error rate.

Coverage uniformity matters as much as average depth. Long-read libraries can have coverage biases caused by fragmentation methods, size selection, or amplification. Regions with low coverage will have poor assembly quality or unreliable variant calls. You should examine the distribution of coverage across the genome, beyond the average, and identify regions that fall below application-specific thresholds.

At a Glance: Key QC Metrics and Interpretation

Practical Workflow for Long-Read QC

Step 1: Collect Raw Signal and Basecall Summary Data

The first QC step begins before basecalling. For Oxford Nanopore data, the raw signal data in POD5 or FAST5 format contains information about pore performance, signal intensity, and base modification signals. For PacBio data, the unaligned BAM file contains the polymerase kinetics and quality information. Tools such as LongReadSum can process these raw formats and produce summary reports that include signal intensity information from the Oxford Nanopore platform [<a href="#ref-1">1</a>].

At this stage, you should check that the instrument produced the expected amount of data, that the signal quality is consistent across the run, and that there are no obvious instrument failures. For Nanopore runs, the number of active pores, the read yield over time, and the quality score distribution over time are the primary metrics. For PacBio runs, the polymerase read length, the number of passes, and the predicted accuracy are the primary metrics.

Step 2: Basecall and Generate Per-Read Metrics

Basecalling converts raw signal into nucleotide sequence with quality scores. The choice of basecaller and its parameters affects read accuracy, read length, and the detection of base modifications. After basecalling, you should generate per-read metrics including read length, mean quality score, and the number of bases above quality thresholds.

For Oxford Nanopore data, the basecaller produces a summary file with per-read statistics. For PacBio HiFi data, the circular consensus sequencing tool produces reads with associated quality scores. You should examine the distribution of these metrics across the entire dataset, beyond the mean values. A dataset with a bimodal quality distribution may indicate that some reads came from a degraded library or that the basecaller performed inconsistently.

Step 3: Screen for Adapters and Contamination

Adapter screening should be performed before assembly or alignment. The screening step identifies reads that contain adapter sequence, reads that are mostly adapter, and reads that contain foreign DNA. For barcoded libraries, you must also verify that barcode assignment is correct.

The fraction of reads with adapter contamination should be reported in the QC summary. If a substantial fraction of reads contain adapters, the library preparation protocol may need to be repeated. If the contamination is limited to specific barcodes, those samples may need to be re-sequenced or the affected reads removed from downstream analysis.

Step 4: Align to Reference or Assemble De Novo

The choice between reference alignment and de novo assembly depends on the application and the availability of a reference genome. For variant calling in a known genome, alignment is the standard approach. For novel genomes or genomes with large structural variation, de novo assembly is required.

Alignment-based QC metrics include mapping rate, the fraction of reads that map uniquely, the distribution of mapping quality scores, and the coverage uniformity across the genome. Assembly-based QC metrics include assembly contiguity (N50, NGA50), completeness (BUSCO), and the number of misassemblies. The Flye assembler, for example, was shown to nearly double the NGA50 of the human genome assembly compared with existing assemblers, demonstrating that assembler choice has a major impact on assembly quality [<a href="#ref-3">3</a>].

Step 5: Evaluate Base Modification Signals

If base modifications are relevant to your study, you should verify that the modification signals are present and correctly called. For Oxford Nanopore data, the basecaller can output modified base calls alongside the canonical base calls. For PacBio HiFi data, methylation can be detected from the kinetics of the polymerase during circular consensus sequencing.

The QC check for base modifications is to compare the fraction of modified bases in your dataset with expected values for your organism and tissue type. Large deviations from expected values may indicate a problem with the library preparation, the basecalling, or the biological sample itself. Tools that summarize base modification information across the dataset are useful for this purpose [<a href="#ref-1">1</a>].

Step 6: Document and Archive QC Results

QC results should be documented in a structured format that allows comparison across runs and across projects. The documentation should include the platform, the basecaller version, the library preparation method, the QC metrics, and any filtering decisions. This documentation is essential for reproducibility and for troubleshooting when downstream analyses fail.

The Galaxy Training Network provides accessible workflow training that emphasizes reproducible analysis practices, and the nf-core documentation describes community standards for pipeline usage and configuration [<a href="#ref-5">5</a>][<a href="#ref-6">6</a>]. Adopting these practices ensures that QC steps are recorded and that the analysis can be repeated by other members of the laboratory.

Tools for Long-Read QC

Cross-Platform QC Tools

LongReadSum is a multi-threaded C++ tool that provides fast and comprehensive QC reports for long-read sequencing data across PacBio, Oxford Nanopore, and Illumina Complete Long Read platforms [<a href="#ref-1">1</a>]. It processes raw signal data, basecall summary files, and unaligned BAM files, and it produces reports covering read, base, base quality, alignment, and base modification metrics. The tool also provides basecalling signal intensity information from the Oxford Nanopore platform, which is useful for diagnosing instrument or library preparation problems.

The advantage of a cross-platform tool is consistency. When you compare QC metrics across platforms, you want to ensure that the metrics are computed in the same way. LongReadSum addresses this need by providing a unified framework for all major long-read platforms [<a href="#ref-1">1</a>].

Platform-Specific Tools

PacBio provides tools for analyzing HiFi and CLR data, including tools for computing read quality, detecting adapters, and evaluating coverage. Oxford Nanopore provides tools for basecalling, demultiplexing, and quality assessment. These platform-specific tools are often the most accurate for platform-specific metrics, but they do not provide cross-platform comparisons.

The choice between cross-platform and platform-specific tools depends on your workflow. If you run a single platform, platform-specific tools may be sufficient. If you run multiple platforms or if you need to compare data across platforms, a cross-platform tool such as LongReadSum is preferable [<a href="#ref-1">1</a>].

Assembly Evaluation Tools

Assembly quality is assessed with tools such as QUAST, which computes assembly statistics including N50, L50, and the number of misassemblies, and BUSCO, which evaluates assembly completeness by searching for conserved single-copy orthologs. The Comprehensive_score (C_score) is a newer metric that combines multiple assembly quality measures [<a href="#ref-4">4</a>].

For lepidopteran genomes assembled from PacBio HiFi data, hifiasm was the preferred assembler, while NextDenovo was preferred for PacBio CLR and ONT data [<a href="#ref-7">7</a>]. The choice of assembler and the evaluation tools should be documented in the QC report.

Common Failure Patterns in Long-Read Sequencing

Low Yield or Short Reads

A common failure pattern is low yield or reads that are much shorter than expected. This can result from degraded DNA, incomplete size selection, or instrument problems. The QC metrics that detect this failure are the read length distribution and the total yield. If the mean read length is far below the expected insert size, the library preparation should be investigated.

For PacBio HiFi sequencing, short reads can result from polymerase molecules that terminate early or from DNA damage that causes the polymerase to fall off. For Oxford Nanopore sequencing, short reads can result from DNA damage, from pores that block or become inactive, or from library preparation issues.

High Error Rates in Specific Sequence Contexts

Another failure pattern is elevated error rates in specific sequence contexts, such as homopolymers, GC-rich regions, or methylated regions. These errors may not be visible in the average quality score but can be detected by examining the error profile across the genome.

For Oxford Nanopore data, homopolymer errors are a known limitation. The basecaller may insert or delete bases in homopolymer regions, and the error rate increases with homopolymer length. For PacBio HiFi data, errors are less common but can occur in regions with extreme GC content or in the presence of base modifications.

Adapter Contamination and Barcode Cross-Talk

Adapter contamination can occur when the library preparation fails to remove adapters completely or when the size selection fails to exclude short adapter-only fragments. Barcode cross-talk can occur when barcodes are misassigned during demultiplexing, leading to contamination between samples.

The QC metrics that detect these failures are the adapter contamination fraction and the barcode assignment accuracy. If a substantial fraction of reads contain adapter sequence, the library preparation should be repeated. If barcode cross-talk is detected, the affected samples should be re-sequenced or the reads should be removed from downstream analysis.

Chimeric Reads and Assembly Artifacts

Chimeric reads can create false connections in assembly graphs and false structural variant calls. They are difficult to detect with standard QC metrics because they align to the reference genome with high confidence. Detection requires either alignment-based methods that look for reads mapping to distant locations or assembly graph inspection.

The Flye assembler constructs a repeat graph from error-prone reads, and the structure of this graph can reveal chimeric connections [<a href="#ref-3">3</a>]. If the assembly graph contains unusual branching patterns or if the assembly contains contigs that do not match the expected genome structure, chimeric reads may be the cause.

Base Modification Signal Loss

Base modification signals can be lost during library preparation, sequencing, or basecalling. The loss can result from DNA damage, from incomplete preservation of modified bases, or from basecaller settings that ignore modification signals.

The QC metric that detects this failure is the fraction of modified bases in the dataset. If the fraction is much lower than expected for the organism and tissue type, the library preparation or basecalling should be investigated. Tools that summarize base modification information across the dataset are useful for this purpose [<a href="#ref-1">1</a>].

Interpreting QC Metrics for Different Applications

Genome Assembly

For genome assembly, the critical QC metrics are read length distribution, coverage depth, and error profile. The read length distribution determines whether reads can span repetitive regions. The coverage depth determines whether the assembler has enough data to resolve the genome structure. The error profile determines which assembler and polisher will work best.

For PacBio HiFi data, more than 20-fold coverage was sufficient for high-quality yeast genome assembly, while ONT data required more than 80-fold coverage [<a href="#ref-4">4</a>]. These thresholds are specific to the organism and the assembler, but they illustrate the general principle that coverage requirements vary by platform and error rate. For lepidopteran genomes, the choice of assembler depended on the platform: hifiasm was preferred for HiFi data, while NextDenovo was preferred for CLR and ONT data [<a href="#ref-7">7</a>].

Structural Variant Detection

For structural variant detection, the critical QC metrics are read length, coverage depth, and the fraction of reads that span variant breakpoints. Structural variants are detected by identifying reads that support non-reference genome structures, such as deletions, insertions, inversions, and duplications. The confidence in a structural variant call depends on the number of supporting reads and the quality of those reads.

Chimeric reads are a particular concern for structural variant detection because they can create false positive calls. The QC workflow should include chimerism detection to ensure that structural variant calls are supported by genuine genomic rearrangements instead of library preparation artifacts.

Transcriptome Analysis

For transcriptome analysis, the critical QC metrics are read length relative to transcript length, coverage uniformity across transcripts, and the fraction of reads that map to the reference transcriptome. Long reads can capture full-length isoforms, but only if the read length distribution matches the transcript length distribution.

For direct mRNA sequencing and cDNA sequencing, the QC metrics include the fraction of reads that contain the poly-A tail, the fraction of reads that span the full-length transcript, and the distribution of read coverage across transcripts. LongReadSum provides use cases for analyzing cDNA sequencing and direct mRNA sequencing data [<a href="#ref-1">1</a>].

Clinical and Diagnostic Applications

For clinical applications, QC metrics determine whether a variant call is reportable or requires orthogonal confirmation. The bioinformatic pipeline and quality metrics, including read length, sequencing depth, and coverage, are critical for accurate variant calling, and validation is often required for variants of uncertain significance or those near detection thresholds [<a href="#ref-2">2</a>].

The QC thresholds for clinical applications are typically more stringent than for research applications. The coverage depth must be sufficient to detect variants at the required sensitivity, and the error rate must be low enough to avoid false positive calls. The QC report should document the metrics that support each variant call.

Records and Documentation for Long-Read QC

What to Record

The QC record should include the platform, the basecaller version, the library preparation method, the sequencing run date, and all QC metrics. The record should also include the filtering decisions and the rationale for those decisions. This documentation is essential for reproducibility and for troubleshooting when downstream analyses fail.

The Galaxy platform's training arm - GTN - offers browser-native tutorials where every step is logged and version-controlled for reproducibility.

How to Use Records for Troubleshooting

When a downstream analysis fails, the QC record can help identify the cause. If the assembly has poor contiguity, the QC record can show whether the read length distribution or coverage depth was inadequate. If the variant calling produces unexpected results, the QC record can show whether the error rate or contamination level was elevated.

The QC record also enables comparison across runs. If a new library preparation protocol produces different QC metrics than the previous protocol, the record can help identify the cause of the difference. This comparison is essential for continuous improvement of laboratory protocols.

Reproducibility Considerations

Reproducibility requires that the QC steps are documented and that the analysis can be repeated. The nf-core documentation describes community standards for pipeline usage and configuration, and the Bioconductor project provides reproducible genomic-analysis documentation [<a href="#ref-8">8</a>][<a href="#ref-6">6</a>]. Adopting these practices ensures that QC steps are recorded and that the analysis can be repeated by other members of the laboratory.

The Carpentries lessons provide foundational training in computing, data, shell, Git, and programming, which is useful for laboratory professionals who need to implement reproducible QC workflows [<a href="#ref-9">9</a>]. The EMBL-EBI Training provides bioinformatics learning pathways and data-resource training [<a href="#ref-10">10</a>].

Limitations of Long-Read QC Metrics

Quality Score Calibration

Quality scores in long-read data are not perfectly calibrated. The estimated error probability may differ from the actual error rate, particularly in specific sequence contexts. This limitation means that quality score thresholds should be validated for each platform and basecaller version.

The practical implication is that you should not rely solely on quality scores for filtering decisions. You should also examine the error profile by aligning reads to a reference genome and comparing the observed errors with the predicted errors.

Platform-Specific Error Profiles

The error profiles of PacBio and Oxford Nanopore data differ significantly. PacBio HiFi reads have high accuracy but can have systematic errors in specific contexts. Oxford Nanopore reads have higher error rates, particularly in homopolymer regions, but the error profile depends on the basecaller version and the pore version.

The practical implication is that QC thresholds must be platform-specific. A quality score threshold that works for PacBio HiFi data may not be appropriate for Oxford Nanopore data. The QC workflow should be validated for each platform and basecaller version.

Reference Bias

Alignment-based QC metrics depend on the reference genome. If the reference genome is incomplete or contains errors, the alignment metrics will be affected. Reads from regions that are absent from the reference genome will not align, and reads from regions with reference errors will have lower mapping quality.

The practical implication is that alignment-based QC metrics should be interpreted with caution when the reference genome is incomplete or when the sample is distantly related to the reference. De novo assembly may be preferable in these cases.

Computational Cost

Long-read QC can be computationally expensive, particularly for large datasets. A single PromethION flowcell can generate terabytes of signal data, which cannot be handled by tools designed for small-scale flowcells [<a href="#ref-1">1</a>]. The QC workflow should be designed to handle the expected data volume, and the computational cost should be considered when selecting tools.

Multi-threaded tools such as LongReadSum are designed to handle large datasets efficiently [<a href="#ref-1">1</a>]. The choice of tools and the computational infrastructure should be documented in the QC record.

Safety and Regulatory Context for Long-Read QC

Clinical Validation Requirements

For clinical applications, QC metrics are part of the validation process. Variants of uncertain significance or those near detection thresholds require validation, and the QC metrics must support the accuracy of the variant calls [<a href="#ref-2">2</a>]. The QC record should document the metrics that support each variant call.

The regulatory requirements for clinical sequencing vary by jurisdiction. Laboratory professionals should be aware of the applicable regulations and should ensure that the QC workflow meets the required standards.

Data Management and Privacy

Long-read sequencing data can contain sensitive genetic information. The QC workflow should include data management practices that protect patient privacy and comply with applicable regulations. The NCBI provides official descriptions of databases, search systems, sequence resources, and analysis services that can support secure data management [<a href="#ref-11">11</a>].

The QC record should document the data management practices, including data storage, access controls, and data sharing agreements.

Professional Escalation Criteria

Laboratory professionals should escalate QC failures to the appropriate personnel when the failure indicates a systemic problem. Escalation criteria include:

• Instrument failures that produce consistently poor quality data
• Library preparation failures that produce adapter contamination or chimeric reads
• Basecalling failures that produce elevated error rates
• Coverage failures that prevent the required sensitivity for variant detection

The escalation should include the QC record, the observed metrics, and the potential causes. The responsible personnel should determine whether the run should be repeated, whether the protocol should be modified, or whether the instrument requires maintenance.

Frequently Asked Questions

What is the difference between read accuracy and quality score in long-read sequencing?

Read accuracy is the fraction of bases in a read that match the true template sequence. Quality score is a predicted probability of error at each base position. The two are related but not identical. Quality scores can be miscalibrated, meaning the predicted error probability differs from the actual error rate. For PacBio HiFi reads, quality scores reflect the consensus accuracy of the circular consensus sequence. For Oxford Nanopore reads, quality scores are produced by the basecaller and depend on the basecaller version and the pore version. You should validate quality score thresholds for each platform and basecaller version by comparing predicted errors with observed errors from reference alignments.

How do I detect chimeric reads in long-read data?

Chimeric reads are reads that contain sequence from two different genomic locations. They can be detected by aligning reads to a reference genome and looking for reads that map to two distant locations with a sharp transition. They can also be detected by examining assembly graphs for unusual branching patterns. The Flye assembler constructs a repeat graph from error-prone reads, and the structure of this graph can reveal chimeric connections [<a href="#ref-3">3</a>]. Chimeric reads are particularly dangerous for genome assembly because they create false connections in the assembly graph, and for structural variant calling because they mimic genuine genomic rearrangements.

What coverage depth is needed for long-read genome assembly?

Coverage requirements depend on the platform and the assembler. For PacBio HiFi genome assembly of yeast, more than 20-fold coverage was sufficient for high-quality assemblies, while ONT data required more than 80-fold coverage [<a href="#ref-4">4</a>]. These thresholds are specific to the organism and the assembler. For lepidopteran genomes, the choice of assembler depended on the platform: hifiasm was preferred for HiFi data, while NextDenovo was preferred for CLR and ONT data [<a href="#ref-7">7</a>]. You should determine coverage requirements empirically for your organism and application.

How do I check for adapter contamination in long-read data?

Adapter contamination is detected by screening reads for adapter sequences using alignment-based or k-mer-based methods. The fraction of reads containing adapter sequence and the fraction of bases that are adapter-derived should be reported. For barcoded libraries, you must also verify that barcode sequences are correctly assigned and that barcode hopping or cross-talk between samples is below acceptable thresholds. If a substantial fraction of reads contain adapters, the library preparation protocol may need to be repeated.

What is the role of base modification signals in long-read QC?

Base modification signals are detected directly from the electrical current signal in Oxford Nanopore sequencing and from the kinetics of the polymerase in PacBio HiFi sequencing. These signals are biologically important for epigenetic studies, but they also affect QC. Base modification signals can influence basecalling accuracy, and the presence of modifications can create systematic errors in certain sequence contexts. Tools that summarize base modification information across the dataset allow you to verify that modification signals are present and that they are distributed as expected [<a href="#ref-1">1</a>].

How do I choose between PacBio HiFi and Oxford Nanopore for my project?

The choice depends on your application, your budget, and your tolerance for error. PacBio HiFi reads have high per-base accuracy and are well suited for variant calling and genome assembly. Oxford Nanopore reads can be very long and are well suited for structural variant detection and for resolving repetitive regions, but they have higher error rates. The QC metrics you need to monitor differ by platform. For PacBio HiFi, you should monitor read length distribution, quality score distribution, and coverage uniformity. For Oxford Nanopore, you should also monitor the error profile in homopolymer regions and the base modification signals.

What tools are available for long-read QC?

LongReadSum is a multi-threaded C++ tool that provides fast and comprehensive QC reports for long-read sequencing data across PacBio, Oxford Nanopore, and Illumina Complete Long Read platforms [<a href="#ref-1">1</a>]. It processes raw signal data, basecall summary files, and unaligned BAM files, and it produces reports covering read, base, base quality, alignment, and base modification metrics. Platform-specific tools are also available from PacBio and Oxford Nanopore. The choice between cross-platform and platform-specific tools depends on your workflow.

How do I document QC results for reproducibility?

Galaxy Training Network courses are structured around the principle that analysis should be portable, auditable, and shareable.

Related Bioinformatics Guides

• Single-Cell RNA Sequencing Quality Control: A Practical Guide to Filtering and Metrics
• Metagenomic Contamination Control: Best Practices for Clean Data
• Long-Read Sequencing Cost and Market: What to Expect
• Long-Read Sequencing for Isoform Quantification: Challenges and Solutions
• RNA-Seq Quality Control: Essential Checks and Tools

References and Further Reading

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