# Systematic Comparison of Error Profiles in PacBio HiFi and Oxford Nanopore Long Reads: Implications for Variant Calling ## Key Takeaways - PacBio HiFi reads exhibit high per-base accuracy dominated by random substitutions, making them amenable to standard variant callers similar to short-read data, though structural variant detection may be limited by read length. - Oxford Nanopore reads, while improving, are characterized by insertion/deletion errors in homopolymers and context-dependent substitutions, necessitating specialized indel-aware polishing and context-aware variant calling tools. - The choice of basecaller version is critical for Oxford Nanopore reproducibility, as different models can alter error profiles and impact downstream variant calling results, requiring strict version pinning. - For structural variant detection, Oxford Nanopore's longer reads offer an advantage in spanning large genomic rearrangements and repetitive elements, whereas PacBio HiFi's accuracy is beneficial for precise breakpoint characterization. - Empirical assessment of error rates and validation of variant calls, particularly in homopolymer regions for Nanopore data, are crucial quality control steps to mitigate common failure patterns like false variants and coverage dropout. - Workflow design for variant calling must account for platform-specific error profiles, influencing coverage depth planning, alignment parameter selection, and the choice of appropriate variant calling algorithms for small and structural variants. --- ## Scope and Reader Context Researchers selecting a long-read sequencing platform for variant detection face a practical problem: PacBio HiFi and Oxford Nanopore reads both produce long molecules, but their error structures differ in ways that materially affect downstream analysis. Substitution errors, insertion and deletion errors, and error distribution across genomic contexts each influence variant caller performance, false positive rates, and the confidence researchers can place in called variants. This article provides a side-by-side statistical breakdown of error types and their genomic context, with direct implications for bioinformatics pipelines used in variant calling. The content is written for biology students, researchers, laboratory professionals, and life-science practitioners who need concrete decision criteria instead of platform marketing comparisons. The central distinction is straightforward. PacBio HiFi reads achieve high per-base accuracy through circular consensus sequencing, producing errors that are largely random substitutions. Oxford Nanopore reads, particularly with newer flow cells and basecalling models, have improved substantially but still show a different error profile dominated by insertion and deletion errors in homopolymer regions and context-dependent substitution patterns. These differences change which variant types each platform can call reliably, how coverage depth should be planned, and which bioinformatics tools are appropriate for each data type. ## At a Glance: Platform Error Profiles and Variant Calling Implications | Error Characteristic | PacBio HiFi | Oxford Nanopore (R10.4.1) | Variant Calling Implication | |---------------------|-------------|--------------------------|----------------------------| | Dominant error type | Random substitutions | Insertions and deletions, context-dependent substitutions | Substitution callers perform well on HiFi, indel-aware polishing needed for Nanopore | | Per-base accuracy | High, typically exceeding Q20 | Lower per-base accuracy, improving with newer chemistry | Higher coverage or polishing required for Nanopore consensus accuracy | | Error distribution | Relatively uniform across sequence context | Concentrated in homopolymers and low-complexity regions | Homopolymer variant calling requires specialized tools or filtering | | Read length | Moderate, typically 10-25 kb | Very long, often exceeding 100 kb | Structural variant detection benefits from Nanopore span, HiFi requires assembly-based approaches | | Methylation detection | Limited native signal | Native signal available without extra steps | Epigenetic studies favor Nanopore native signal | | Basecalling dependency | Consensus-based, less dependent on basecaller | Highly dependent on basecaller model and version | Pipeline reproducibility requires pinned basecaller versions | This comparison reflects the current state of both platforms as documented in benchmarking literature. The R10.4.1 flow cell chemistry has narrowed but not eliminated the accuracy gap with short-read platforms, and the choice between platforms should be framed as an important part of study design instead of a search for a single best option. ## Error Mechanisms and Their Genomic Context ### PacBio HiFi Circular Consensus Sequencing PacBio HiFi reads are produced by sequencing the same circular template multiple times and generating a consensus sequence from the subreads. This circular consensus approach averages out random errors that occur during individual passes, resulting in high per-base accuracy. The error profile that remains after consensus generation is dominated by random substitutions that are largely independent of sequence context. The practical consequence for variant calling is that HiFi data behaves similarly to short-read data in terms of error structure, though with much longer read lengths. Most short-read variant callers can be adapted to HiFi data with minimal modification, and the substitution errors that remain are handled well by standard statistical models that assume independent base errors. For structural variant detection, HiFi reads provide sufficient length to span many repetitive elements and segmental duplications, though the upper limit of read length is lower than what Nanopore can achieve. Researchers working on complex genomes with large structural variants should consider whether the read length distribution of HiFi is adequate for their specific variant classes of interest. ### Oxford Nanopore Native Sequencing and Basecalling Oxford Nanopore sequencing measures changes in ionic current as DNA passes through a protein nanopore. The raw signal is converted to nucleotide sequence through basecalling algorithms that have improved substantially across chemistry and software versions. The R10.4.1 flow cell chemistry has narrowed the accuracy gap with short-read platforms, but systematic biases remain. The error profile of Nanopore reads is characterized by insertion and deletion errors, particularly in homopolymer regions where the signal difference between consecutive identical bases is small. Substitution errors also occur and are context-dependent, with certain motifs showing higher error rates than others. These context-dependent errors are problematic for variant calling because they can create systematic false positives at specific sequence positions. The dependency on basecalling models introduces a reproducibility consideration. Different basecaller versions can produce different error profiles from the same raw signal data, which means that variant calling results may vary across software versions even when the underlying sequencing data is identical. Researchers should pin basecaller versions and document them in analysis pipelines to ensure reproducibility. ### Error Context and Variant Caller Performance The genomic context of errors matters as much as the overall error rate. A platform with low average error but concentrated errors in specific contexts can perform worse for variant calling than a platform with higher average error that is uniformly distributed. This is because variant callers model error probabilistically, and systematic errors violate the assumptions of these models. For PacBio HiFi, the relatively uniform error distribution means that variant callers can confidently identify true variants as deviations from the expected error model. For Oxford Nanopore, the concentration of errors in homopolymers and low-complexity regions means that variants in these contexts are called with lower confidence, and false positives in these regions are more common. Researchers planning variant detection studies should assess the sequence context of their target regions. If the genomic regions of interest contain long homopolymers or low-complexity sequence, Nanopore data will require additional filtering or specialized variant callers that model context-dependent errors. If the target regions are relatively complex and unique, both platforms can perform well with appropriate analysis approaches. ## Workflow Design for Variant Calling ### Data Inputs and Quality Assessment The first step in any long-read variant calling workflow is quality assessment of the raw sequencing data. For both platforms, this involves examining read length distributions, per-base quality scores, and total yield relative to the genome size and desired coverage. For PacBio HiFi data, quality assessment typically focuses on the accuracy of the consensus reads, which is reflected in the predicted accuracy scores assigned to each read. Reads with lower predicted accuracy should be filtered or flagged for downstream analysis. The read length distribution is also important, as very short reads may not provide sufficient information for confident variant calling. For Oxford Nanopore data, quality assessment is more complex because the basecaller assigns quality scores that may not perfectly reflect true error rates. Researchers should examine the distribution of quality scores, the read length distribution, and, if possible, compare a subset of reads to a known reference to estimate empirical error rates. The choice of basecaller model and version should be documented at this stage. The [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible workflow training and analysis tutorials that cover quality assessment and downstream analysis for various sequencing platforms. These resources are useful for researchers who are new to long-read analysis or who want to standardize their workflows. ### Alignment and Mapping Considerations Read alignment is a critical step that influences variant calling accuracy. Long reads require aligners that can handle the error profiles of each platform. For PacBio HiFi data, aligners designed for accurate long reads work well because the error profile is similar to short reads. For Oxford Nanopore data, aligners must account for the higher indel rates and the context-dependent nature of errors. The choice of alignment parameters affects variant calling outcomes. Researchers should consider whether to use splice-aware alignment for transcriptome data, whether to allow soft clipping at read ends, and how to handle reads that map to multiple locations. Multi-mapping reads are particularly problematic for variant calling because they can create false positive variant calls at repetitive regions. For structural variant detection, the alignment approach differs from small variant calling. Structural variant callers often use split-read or assembly-based approaches that require different alignment strategies. Researchers should select alignment parameters based on the primary variant type of interest and document these choices for reproducibility. ### Variant Calling Tools and Parameters The choice of variant caller depends on both the platform and the variant type of interest. For small variants (single nucleotide variants and small indels), different tools have been optimized for different error profiles. Tools designed for PacBio HiFi data typically model substitution errors and handle the high accuracy of the reads. Tools designed for Oxford Nanopore data must model the higher indel rates and context-dependent errors. For structural variants, the approach differs. Some tools use read depth and split-read information, while others use assembly-based approaches that compare the assembled genome to a reference. The choice of approach depends on the size and type of structural variants of interest, as well as the read length distribution of the platform. The [Bioconductor](https://bioconductor.org/) project provides official package documentation and reproducible genomic-analysis workflows that include variant calling tools and associated quality control steps. Researchers should consult these resources when selecting and configuring variant calling tools for their specific data types. ### Coverage Depth Planning Coverage depth requirements differ between platforms and variant types. For PacBio HiFi data, the high per-base accuracy means that lower coverage can be sufficient for small variant calling, though higher coverage improves confidence in heterozygous variant calls and reduces false negatives. For Oxford Nanopore data, the higher error rate means that higher coverage is typically required to achieve comparable variant calling accuracy. The relationship between coverage and variant calling accuracy is not linear. Increasing coverage from 10x to 20x typically produces substantial improvements in variant calling accuracy, while increasing from 30x to 40x produces smaller gains. Researchers should consider the expected allele frequencies of the variants they are studying, as low-frequency variants require higher coverage to detect with confidence. For structural variant detection, coverage requirements depend on the size and type of variants. Large deletions and insertions can be detected at lower coverage because they create clear patterns in read depth and alignment. Smaller structural variants and those in repetitive regions require higher coverage and longer reads for confident detection. ## Options and Tradeoffs Across Use Cases ### Small Variant Detection For single nucleotide variant and small indel detection, PacBio HiFi data generally provides higher accuracy than Oxford Nanopore data at equivalent coverage. The random substitution error profile of HiFi is well handled by standard variant calling models, and the high per-base accuracy reduces false positive rates. Oxford Nanopore data can achieve comparable small variant calling accuracy with sufficient coverage and appropriate analysis tools, but the context-dependent error profile requires more careful filtering and validation. Variants in homopolymer regions and low-complexity sequence should be treated with caution regardless of the platform used. Researchers working on clinical samples or other applications where false positive variants have significant consequences should consider the higher accuracy of PacBio HiFi for small variant detection. The cost difference between platforms may be justified by the reduced need for validation of called variants. ### Structural Variant Detection For structural variant detection, Oxford Nanopore reads have an advantage in read length, which allows individual reads to span large structural variants and repetitive elements. This is particularly valuable for detecting large insertions, deletions, inversions, and translocations that are difficult to resolve with shorter reads. PacBio HiFi reads can also detect structural variants, particularly through assembly-based approaches, but the shorter read length means that some large variants may not be fully spanned by individual reads. The higher accuracy of HiFi reads can be advantageous for resolving breakpoints and characterizing the precise sequence of structural variant junctions. The choice between platforms for structural variant detection depends on the size distribution of variants of interest and the complexity of the genomic region. For very large variants in complex regions, the longer reads of Oxford Nanopore provide an advantage. For precise breakpoint characterization, the higher accuracy of PacBio HiFi may be preferable. ### Metagenomic and Microbial Applications Long-read sequencing has particular advantages for microbial applications, including the ability to reconstruct complete bacterial genomes and to detect antimicrobial resistance determinants. The [comparative evaluation of sequencing technologies for detecting antimicrobial resistance in bloodstream infections](https://doi.org/10.3390/antibiotics14121257) highlights that sequencing-based diagnostics offer measurable improvements in sensitivity and turnaround time compared to culture-based approaches. For bacterial genome reconstruction, both platforms can produce complete genomes, but the error profiles affect the accuracy of the final assembly. PacBio HiFi data can produce highly accurate assemblies with relatively simple workflows, while Oxford Nanopore data may require additional polishing steps to achieve comparable accuracy. The [StrainCascade workflow](https://doi.org/10.1016/j.isci.2026.116189) demonstrates an automated, modular approach to long-read bacterial genome reconstruction that integrates assembly, annotation, and functional profiling in a reproducible framework. The choice of platform for microbial applications depends on the specific research questions. For antimicrobial resistance gene detection, the accuracy of the assembly matters because resistance genes can be located in repetitive or mobile genetic elements. For outbreak investigations, the speed of Oxford Nanopore sequencing may be advantageous despite the lower per-base accuracy. ### Transcriptome and Isoform Quantification Long-read sequencing has transformed transcriptome analysis by enabling the direct observation of full-length isoforms. The [lr-kallisto approach](https://doi.org/10.1371/journal.pcbi.1013692) demonstrates that fast and accurate quantification of long-read data is possible, and that it is improved by exome capture. For transcriptome applications, the error profile of the sequencing platform affects isoform quantification accuracy. Errors in the reads can lead to incorrect assignment of reads to isoforms, particularly for isoforms that differ by small indels or single nucleotide variants. The higher accuracy of PacBio HiFi data may be advantageous for distinguishing closely related isoforms, while the longer reads of Oxford Nanopore may be advantageous for spanning complex splicing events. The choice of platform for transcriptome analysis should consider the isoform complexity of the organism and the specific quantification questions. Researchers should also consider whether native RNA sequencing (available on Oxford Nanopore) or cDNA sequencing is more appropriate for their applications, as native RNA sequencing preserves base modifications but may have different error characteristics. ### Genome Assembly and Reference Generation Both platforms are used for genome assembly, often in combination with each other and with short-read data. The [chromosome-scale reference genome assembly of tetraploid sainfoin](https://doi.org/10.1007/s00425-026-05021-y) demonstrates the use of PacBio HiFi, Oxford Nanopore, Illumina short read, and Hi-C data to produce a haplotype-resolved, chromosome-scale assembly. For genome assembly, the complementary strengths of the two platforms are often exploited. PacBio HiFi data provides accurate sequence for the assembly, while Oxford Nanopore data provides long reads that span repetitive regions and resolve structural complexity. The combination of both platforms can produce assemblies that are more contiguous and accurate than either platform alone. Researchers planning genome assembly projects should consider the tradeoff between the cost of using multiple platforms and the improved assembly quality. For complex genomes with high repeat content or polyploidy, the combination of platforms may be necessary to achieve reference-grade assemblies. ## Observations and Measurements ### Empirical Error Rate Assessment Before committing to a large-scale variant calling project, researchers should empirically assess the error rates of their sequencing data. This can be done by sequencing a well-characterized reference sample and comparing the called variants to the known reference sequence. For this assessment, researchers should calculate the false positive and false negative rates for different variant types and in different sequence contexts. This information is essential for interpreting variant calls in the actual study samples and for setting appropriate filtering thresholds. The empirical error rate assessment should be repeated when changing sequencing platforms, chemistries, basecaller versions, or analysis pipelines. Error profiles can change with any of these variables, and assumptions based on previous data may not hold for new data. ### Basecaller Version Effects For Oxford Nanopore data, the basecaller version has a substantial effect on error profiles. Newer basecaller models typically produce more accurate reads, but they may also change the distribution of errors across sequence contexts. Researchers should document the basecaller version used for each dataset and consider whether results from different basecaller versions can be directly compared. The effect of basecaller version on variant calling accuracy should be evaluated empirically. Researchers can basecall the same raw data with different versions and compare the resulting variant calls. This comparison can reveal whether apparent differences between samples are due to biological variation or to basecaller version differences. For PacBio HiFi data, the consensus approach reduces the dependency on basecaller versions, but the underlying sequencing chemistry and software versions can still affect error profiles. Researchers should document all software versions used in the sequencing and analysis pipeline. ### Coverage and Error Rate Interactions The relationship between coverage and variant calling accuracy depends on the error rate of the platform. For platforms with higher error rates, higher coverage is needed to achieve the same variant calling confidence. This interaction should be considered when planning sequencing depth. Researchers can estimate the coverage needed for their specific application by simulating or empirically testing different coverage levels. This can be done by subsampling existing data to lower coverage levels and assessing the effect on variant calling accuracy. The coverage needed also depends on the allele frequency of the variants of interest. Rare variants require higher coverage to be detected with confidence, and the required coverage increases as the allele frequency decreases. ## Records and Documentation ### Metadata Requirements Reproducible variant calling requires comprehensive metadata documentation. For each sequencing dataset, researchers should record the platform, chemistry version, flow cell version, basecaller version, and all analysis software versions. This information is essential for interpreting results and for reproducing analyses. The [nf-core documentation](https://nf-co.re/docs) provides standards for community pipeline usage and configuration that emphasize reproducibility. Researchers should adopt similar standards for their own workflows, documenting all parameters and versions used in the analysis. Metadata should be stored in a structured format that can be queried and analyzed. This allows researchers to assess the effects of different sequencing and analysis parameters on variant calling results across multiple datasets. ### Analysis Pipeline Documentation The analysis pipeline should be documented in sufficient detail that another researcher could reproduce the analysis. This includes the exact commands used, the parameters for each tool, and the versions of all software. Containerization or workflow management systems can help ensure reproducibility. Researchers should also document the rationale for parameter choices. This is important because parameters that work well for one data type or organism may not be appropriate for another. The documentation should explain why specific parameters were chosen and what alternatives were considered. The [The Carpentries lessons](https://carpentries.org/lessons) provide foundational training in computing, data, shell, Git, and programming that supports reproducible analysis practices. Researchers who are new to these practices should consider completing relevant lessons before starting their analysis. ### Variant Call Records Variant calls should be recorded in standard formats that include quality scores and filtering information. The records should distinguish between high-confidence and low-confidence variant calls, and should include information about the evidence supporting each call. For each variant call, researchers should record the read depth, the number of reads supporting the variant allele, the quality scores, and the filtering criteria applied. This information is essential for interpreting the biological significance of the variant and for comparing results across samples. Variant call records should also include information about the genomic context of the variant, such as whether it falls in a repetitive region, a homopolymer, or a low-complexity region. This context information is important for assessing the confidence of the variant call. ## Quality Controls and Validation ### Read-Level Quality Filters Read-level quality filters are the first line of defense against false variant calls. For PacBio HiFi data, filters based on predicted accuracy can remove low-quality reads that might introduce errors. For Oxford Nanopore data, filters based on read length and quality scores can remove short or low-quality reads that are more likely to contain errors. The choice of filtering thresholds should be based on the empirical error rate assessment and the specific requirements of the variant calling application. Stringent filters reduce false positives but may also reduce sensitivity, particularly for variants in low-complexity regions. Researchers should assess the effect of different filtering thresholds on variant calling results. This can be done by running the analysis with different thresholds and comparing the resulting variant calls. ### Alignment-Based Quality Controls Alignment-based quality controls assess the quality of the read alignment and identify potential problems. These include checking the mapping quality distribution, the coverage uniformity, and the presence of alignment artifacts. For Oxford Nanopore data, alignment-based quality controls are particularly important because the higher error rate can lead to misalignment in repetitive regions. Researchers should examine the alignment patterns in these regions and consider whether additional filtering or realignment is needed. For PacBio HiFi data, alignment-based quality controls are less critical but still important. The high accuracy of the reads means that misalignment is less common, but it can still occur in complex genomic regions. ### Variant-Level Validation Variant-level validation is essential for confirming the accuracy of variant calls. This can involve comparing variant calls to known reference samples, validating variants with an orthogonal method such as Sanger sequencing or short-read sequencing, or using statistical approaches to assess the confidence of variant calls. The level of validation needed depends on the application. For research applications, statistical validation may be sufficient. For clinical applications, orthogonal validation is typically required. Researchers should establish validation criteria before starting the analysis and document the validation results for each variant call. This documentation is essential for interpreting the biological significance of the variants and for publication. ## Common Failure Patterns ### Homopolymer-Associated False Variants The most common failure pattern in Oxford Nanopore variant calling is the presence of false variants in homopolymer regions. These false variants are caused by the difficulty of accurately determining the length of homopolymers from the nanopore signal. Researchers should be particularly cautious about variant calls in homopolymer regions, especially when the variant is an insertion or deletion. These calls should be validated with an orthogonal method before being considered reliable. For PacBio HiFi data, homopolymer-associated false variants are less common but can still occur. The circular consensus approach reduces but does not eliminate the difficulty of accurately sequencing homopolymers. ### Coverage Dropout in Repetitive Regions Coverage dropout in repetitive regions is a common problem for both platforms. Reads that map to multiple locations in the genome are often filtered out or assigned low mapping quality, leading to reduced coverage in these regions. The reduced coverage in repetitive regions can lead to false negative variant calls, particularly for heterozygous variants. Researchers should assess coverage in repetitive regions and consider whether additional sequencing or alternative analysis approaches are needed. For structural variant detection, coverage dropout in repetitive regions is particularly problematic because structural variants often occur in or near repetitive elements. Researchers should be cautious about interpreting the absence of structural variants in these regions. ### Basecaller Version Inconsistency Basecaller version inconsistency is a common failure pattern in Oxford Nanopore analysis. If different samples in a study are basecalled with different versions, the error profiles may differ, leading to apparent differences in variant calls that are actually due to technical variation. Researchers should basecall all samples in a study with the same basecaller version and document the version used. If basecaller versions must be changed, the effect on variant calling should be assessed empirically. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide learning pathways for bioinformatics data-resource training and practical analysis education that can help researchers understand and avoid these failure patterns. ### Reference Bias in Variant Calling Reference bias occurs when the reference genome used for alignment influences the variant calls. This can happen when reads that differ from the reference are more likely to be misaligned or filtered out, leading to false negative variant calls. Reference bias is a particular concern for Oxford Nanopore data because the higher error rate can make reads that differ from the reference more difficult to align. Researchers should consider using a reference that is closely related to the sample being studied, or using reference-free approaches such as assembly-based variant calling. For PacBio HiFi data, reference bias is less of a concern because the higher accuracy of the reads makes alignment more robust. However, reference bias can still occur in repetitive regions or when the sample is distantly related to the reference. ## Limitations and Interpretation Boundaries ### Platform-Specific Limitations Each platform has limitations that affect the interpretation of variant calling results. For PacBio HiFi, the read length is limited compared to Oxford Nanopore, which can affect the detection of very large structural variants. The cost per base is also typically higher than Oxford Nanopore. For Oxford Nanopore, the higher error rate and context-dependent error profile require more careful analysis and validation. The dependency on basecalling models introduces a reproducibility consideration that is less prominent for PacBio HiFi. Researchers should understand these limitations and design their studies accordingly. The choice of platform should be based on the specific research questions and the tradeoffs between accuracy, read length, cost, and turnaround time. ### Context-Dependent Interpretation Variant calling results should be interpreted in the context of the sequencing platform and analysis pipeline used. Variants that are called with high confidence on one platform may not be called on another platform, and the biological significance of a variant should be assessed independently of the platform used to detect it. The [review of sequencing methods for soil microbiome studies](https://doi.org/10.3390/microorganisms14051132) demonstrates that method choice has a strong effect on which species are detected and how the community is described. This principle applies to variant calling as well: the choice of platform and analysis pipeline affects which variants are detected and how they are characterized. Researchers should report the platform and analysis details in publications and should be cautious about comparing variant calls across studies that used different platforms or pipelines. ### Generalizability Boundaries The error profiles described in this article are based on current platform versions and chemistries. Both platforms are evolving rapidly, and future versions may have different error profiles. Researchers should not assume that the error profiles described here apply to future platform versions. The error profiles also vary across organisms and sample types. The performance of a platform for variant calling in one organism may not predict its performance in another organism with different genome complexity or base composition. Researchers should empirically assess error profiles for their specific application instead of relying solely on published benchmarks. This is particularly important for novel organisms or unusual sample types. ## Safety and Regulatory Context ### Clinical and Diagnostic Applications For clinical and diagnostic applications, the choice of sequencing platform has regulatory implications. The accuracy of variant calls is critical for patient care, and false positive or false negative variants can have serious consequences. The [comparative evaluation of sequencing technologies for detecting antimicrobial resistance in bloodstream infections](https://doi.org/10.3390/antibiotics14121257) highlights the importance of context-specific strategies in clinical applications. Sequencing-based diagnostics offer measurable improvements in sensitivity and turnaround time, but the choice of platform and analysis approach must be tailored to the clinical question. Researchers and clinicians should be aware of the regulatory requirements for sequencing-based diagnostics in their jurisdiction. The validation requirements for clinical variant calling are typically more stringent than for research applications. ### Data Privacy and Security Sequencing data can contain sensitive information about individuals, and researchers must ensure that data is stored and processed securely. This is particularly important for clinical samples and for data that will be shared or deposited in public databases. The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) provide official descriptions of databases, search systems, sequence resources, and analysis services that include data submission and access policies. Researchers should be aware of the data sharing requirements for their funding sources and journals. Researchers should also be aware of the privacy implications of sharing raw sequencing data, which can be used to identify individuals even when identifying information is removed. ### Professional Escalation Criteria Researchers should escalate to professional bioinformaticians or sequencing facility staff when they encounter problems that exceed their expertise. This includes persistent quality issues, unexpected error profiles, or analysis results that are inconsistent with biological expectations. Specific escalation criteria include: coverage that is substantially lower than expected, error rates that are substantially higher than expected for the platform and chemistry, variant calls that are inconsistent across replicates, and analysis pipeline failures that cannot be resolved with standard troubleshooting. Researchers should also escalate when they are considering changing platforms, chemistries, or analysis approaches for a large-scale study. The cost of switching platforms or pipelines is substantial, and professional guidance can help avoid costly mistakes. ## Frequently Asked Questions ### How do PacBio HiFi and Oxford Nanopore error profiles differ for small variant calling? PacBio HiFi reads have a relatively uniform error profile dominated by random substitutions, which is well handled by standard variant calling models. Oxford Nanopore reads have higher error rates with a context-dependent profile, including more insertion and deletion errors in homopolymer regions. For small variant calling, PacBio HiFi generally provides higher accuracy at equivalent coverage, but Oxford Nanopore can achieve comparable accuracy with sufficient coverage and appropriate analysis tools. ### What coverage depth is recommended for variant calling with each platform? Coverage requirements depend on the variant type, allele frequency, and genomic context. PacBio HiFi data typically requires lower coverage than Oxford Nanopore data for equivalent small variant calling accuracy because of the higher per-base accuracy. Researchers should empirically assess coverage requirements for their specific application by subsampling data and assessing the effect on variant calling accuracy. ### How should homopolymer regions be handled in Oxford Nanopore variant calling? Homopolymer regions are a known source of false variants in Oxford Nanopore data because of the difficulty of accurately determining homopolymer length from the nanopore signal. Researchers should treat variant calls in homopolymer regions with caution, particularly insertion and deletion variants. Validation with an orthogonal method is recommended before considering these variants reliable. ### Can PacBio HiFi and Oxford Nanopore data be combined in a single analysis? Yes, combining data from both platforms can be advantageous for certain applications. The complementary strengths of the two platforms can be exploited, with PacBio HiFi providing accurate sequence and Oxford Nanopore providing long reads that span repetitive regions. The [chromosome-scale reference genome assembly of tetraploid sainfoin](https://doi.org/10.1007/s00425-026-05021-y) demonstrates the successful combination of both platforms with short-read and Hi-C data. ### How does basecaller version affect Oxford Nanopore variant calling? Basecaller version has a substantial effect on Oxford Nanopore error profiles. Newer basecaller models typically produce more accurate reads, but they may also change the distribution of errors across sequence contexts. Researchers should basecall all samples in a study with the same basecaller version and document the version used. The effect of basecaller version on variant calling should be assessed empirically. ### What is the best approach for structural variant detection with long reads? The best approach depends on the size and type of structural variants of interest. Oxford Nanopore reads have an advantage in read length, which allows individual reads to span large structural variants. PacBio HiFi reads can also detect structural variants, particularly through assembly-based approaches, and the higher accuracy can be advantageous for resolving breakpoints. Researchers should consider the size distribution of variants of interest and the complexity of the genomic region. ### How should variant calling results be validated? Variant calling results should be validated using multiple approaches. This includes comparing variant calls to known reference samples, validating variants with an orthogonal method such as Sanger sequencing or short-read sequencing, and using statistical approaches to assess the confidence of variant calls. The level of validation needed depends on the application, with clinical applications requiring more stringent validation than research applications. ### What are the main considerations for choosing between PacBio HiFi and Oxford Nanopore for a variant calling study? The main considerations are the variant types of interest, the genomic context of the target regions, the required accuracy, the read length needed, the cost, and the turnaround time. PacBio HiFi provides higher per-base accuracy and a more uniform error profile, while Oxford Nanopore provides longer reads and native signal for methylation detection. The choice should be based on the specific research questions and the tradeoffs between these factors. ## Related Bioinformatics Guides - [How to Choose a Long-Read Sequencing Platform: PacBio vs Oxford Nanopore](/knowledge/bioinformatics/how-to-choose-a-long-read-sequencing-platform-pacbio-vs-oxford-nanopore) - [Long-Read Metagenome Assembly: Overcoming Challenges with Nanopore and PacBio Data](/knowledge/bioinformatics/long-read-metagenome-assembly-overcoming-challenges-with-nanopore-and-pacbio-data) - [Long-Read Sequencing Technologies: PacBio and Oxford Nanopore](/knowledge/bioinformatics/long-read-sequencing-technologies-pacbio-and-oxford-nanopore) - [Detecting Structural Variants with Long-Read Sequencing: Methods and Considerations](/knowledge/bioinformatics/detecting-structural-variants-with-long-read-sequencing-methods-and-considerations) - [Metagenomics Pipeline: From Raw Reads to Taxonomic and Functional Profiles](/knowledge/bioinformatics/metagenomics-pipeline-from-raw-reads-to-taxonomic-and-functional-profiles) ## Related Clinical & Scientific Guides * [A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data](/knowledge/bioinformatics/a-practical-guide-to-detecting-antimicrobial-resistance-genes-in-shotgun-metagenomic-data) * [Computational Immunology: Modeling the Immune System](/knowledge/bioinformatics/computational-immunology-modeling-the-immune-system) * [How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices](/knowledge/bioinformatics/how-to-set-hard-filters-for-germline-variant-calling-a-practical-guide-to-gatk-best-practices) ## References and Further Reading - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Bioconductor](https://bioconductor.org/). Bioconductor Project. - [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project. - [nf-core Documentation](https://nf-co.re/docs). nf-core. - [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries. - [Choosing Between Short-Read 16S, Full-Length ONT 16S, and Long-Read Shotgun Metagenomics for Soil Microbiome Studies: A Critical Review of the Benchmarking Evidence.](https://doi.org/10.3390/microorganisms14051132). 2026. - [A chromosome-scale reference genome and integrative transcriptome provide insight into tissue- and stress-specific responses in tetraploid sainfoin (Onobrychis viciifolia).](https://doi.org/10.1007/s00425-026-05021-y). 2026. - [Long-read sequencing transcriptome quantification with lr-kallisto.](https://doi.org/10.1371/journal.pcbi.1013692). 2025. - [Comparative Evaluation of Sequencing Technologies for Detecting Antimicrobial Resistance in Bloodstream Infections.](https://doi.org/10.3390/antibiotics14121257). 2025. - [<,i>,StrainCascade<,/i>,: An automated, modular workflow for high-throughput long-read bacterial genome reconstruction and characterization.](https://doi.org/10.1016/j.isci.2026.116189). 2026. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.