# Phasing Structural Variants with Long Reads: How to Resolve Allele-Specific SV Calls Structural variants (SVs) are large genomic rearrangements that often remain unresolved in standard variant calling workflows. When a researcher identifies an SV in long-read sequencing data, the next question is which parental chromosome carries the variant and how that allele-specific arrangement affects gene expression. This article describes a practical workflow for phasing SVs using long-read alignments, integrating SV calls with SNP-based haplotypes, and producing allele-specific SV calls that can be used for downstream functional analysis. The workflow presented here applies to researchers working with Oxford Nanopore Technologies (ONT) data, Pacific Biosciences (PacBio) HiFi data, or a combination of both. The methods assume familiarity with basic command-line operations and alignment of long reads to a reference genome. For researchers who need to build these foundational skills, training materials from [The Carpentries](https://carpentries.org/lessons) cover shell scripting, data organization, and reproducible computing practices that support complex genomic analyses. ## The Problem of Unphased Structural Variants A structural variant call from a standard long-read pipeline tells you that a deletion, insertion, inversion, or duplication exists at a genomic coordinate. It does not tell you which homologous chromosome carries that variant. In a diploid genome, each position has two alleles, one inherited from each parent. When a deletion spans a gene promoter, the consequence depends entirely on whether the deletion is on the same chromosome as a regulatory variant or on the opposite chromosome. Unphased SV calls create several practical problems. First, you cannot determine whether two SVs in the same genomic region are on the same haplotype or on different haplotypes. This distinction matters for understanding compound effects. Second, you cannot connect an SV to expression data without knowing which allele is expressed. Third, you cannot determine whether a variant is inherited from the mother or the father without phasing against parental genotypes or using pedigree information. The solution is to phase the SV calls. Phasing assigns each heterozygous variant to one of two haplotypes. For small variants, phasing tools have existed for years. For structural variants, the problem is more difficult because SVs are large, often span repetitive regions, and may not be captured by the same read pairs that phase small variants. Long reads solve part of this problem because they are long enough to span entire SVs and the flanking SNPs that anchor them to a haplotype. A single PacBio HiFi read of 13.5 kilobases average length can cover a small deletion and several surrounding SNPs, providing direct evidence that the deletion and those SNPs are on the same molecule. The challenge is integrating this information into a coherent haplotype across the genome. ## What Long-Read Platforms Offer for Phasing The two dominant long-read platforms take different approaches to accuracy and read length, and these differences affect phasing strategy. PacBio HiFi sequencing produces reads that are both long and highly accurate. The circular consensus sequencing approach reads the same molecule multiple times and produces a consensus sequence with reported accuracy of 99.8% and average read length of 13.5 kilobases. In the human HG002 genome, this approach achieved precision and recall rates of at least 99.91% for single-nucleotide variants, 95.98% for small insertions and deletions, and 95.99% for structural variants. Nearly all variants, 99.64%, could be phased into haplotypes. The combination of length and accuracy means that a single HiFi read can provide reliable evidence for both the SV and the flanking SNPs that determine its haplotype assignment. Oxford Nanopore sequencing produces reads that can be much longer than HiFi reads, but with lower per-base accuracy in standard mode. Ultra-long reads from the MinION platform have reached N50 values exceeding 100 kilobases with individual reads up to 882 kilobases. These very long reads can span entire genes and complex rearrangements. In the GM12878 human cell line, ultra-long reads enabled assembly and phasing of the 4-megabase major histocompatibility complex locus in its entirety. The tradeoff is that lower base accuracy makes SNP detection within a single read less reliable, so phasing may require either higher coverage or integration with short-read SNP calls. For phasing structural variants, the choice between platforms depends on the variant types you need to resolve. HiFi reads provide accurate SNP calls within the same read that carries the SV, which simplifies direct phasing. Ultra-long ONT reads provide haplotype continuity across very large regions, which helps connect distant SVs to the same parental chromosome. ## At a Glance: Phasing Workflow Options The table below summarizes the main workflow options for phasing structural variants with long-read data. Each option makes different assumptions about input data and produces different output formats. | Workflow Option | Input Data | Phasing Method | Best For | Output | | --- | --- | --- | --- | --- | | Joint phasing with HiPhase | PacBio HiFi reads, SNV calls, SV calls | Read-based phasing using A* algorithm | Resolving SNVs, indels, SVs, and tandem repeats together | Phased VCF with allele assignments for all variant types | | Pre-phasing with SVUPP | Long reads (ONT or HiFi), SV calls, per-read phasing from WhatsHap or QUILT2 | Incorporates read phasing into genotype likelihoods | Improving SV genotyping accuracy when SVs are not near other SVs | Genotyped SVs with phasing-aware likelihood scores | | Reference-panel phasing | Long reads, SNP genotypes, reference panel | QUILT2 or similar reference-based phasing | Populations with available reference panels | Phased SNP haplotypes for anchoring SVs | | Assembly-based phasing | Ultra-long reads, de novo assembly | Haplotype-resolved assembly | Complex SVs in repetitive regions, MHC locus, telomere-adjacent variants | Phased contigs with SV annotations | The choice of workflow depends on your research question. If you need to phase all variant types in a single sample and you have HiFi data, joint phasing with HiPhase is the most direct approach. If you have already called SVs and need to improve genotyping accuracy, the SVUPP approach can be applied to existing calls. If you are working with a population sample and have a reference panel, reference-based phasing may provide more complete haplotypes. ## Core Principles of SV Phasing ### Haplotype Assignment Requires Heterozygous Anchors Phasing works by connecting a variant of interest to nearby heterozygous SNPs that have already been assigned to haplotypes. A heterozygous SNP has two alleles, one on each chromosome. If a long read carries both the SV and one allele of a heterozygous SNP, then the SV and that SNP allele are on the same molecule, and therefore on the same haplotype. The practical requirement is that your data must contain heterozygous SNPs near the SV. If the region around the SV is homozygous for all SNPs, there is no information to distinguish the two haplotypes, and the SV cannot be phased regardless of read length. This situation is common in regions of low genetic diversity or in samples with high relatedness. ### Read Length Determines Phasing Continuity A read can only phase variants that it physically spans. If an SV is 50 kilobases long and your reads are 15 kilobases, no single read will span the entire SV and both flanking regions. You may still phase the SV if reads cover the breakpoints and connect to SNPs on either side, but the phasing will be less direct. Ultra-long reads provide a solution for very large SVs. The ability to generate reads exceeding 100 kilobases means that even large deletions and inversions can be spanned by a single molecule. This approach was essential for resolving the MHC locus, which contains complex structural variation that confounds shorter reads. ### Phasing Blocks Have Boundaries Phasing does not produce a complete haplotype for the entire genome in one step. Instead, it produces phase blocks, which are contiguous genomic regions where variants have been assigned to haplotypes. The boundaries between blocks occur where there is insufficient information to connect the haplotypes across the gap. The phase block NG50 is a useful metric for evaluating phasing quality. HiPhase produced an average phase block NG50 of 480 kilobases with 929 switchflip errors in its assessment. A switchflip error occurs when the phasing switches from one haplotype to the other within a block, meaning that variants on one side of the switch are assigned to the wrong haplotype relative to variants on the other side. For SV phasing, the relevant question is whether the SV falls within a phase block that contains enough SNP anchors to assign it confidently. If the SV is near a block boundary, you may need additional data or a different phasing strategy. ## Preparing Input Data for SV Phasing ### Alignment Quality Controls Before attempting to phase SVs, you need high-quality alignments of your long reads to a reference genome. Poor alignments produce false SV calls and incorrect SNP genotypes, both of which corrupt phasing. For PacBio HiFi data, align reads with a mapper designed for accurate long reads. Check the alignment statistics: mapping rate, median read length after mapping, and coverage uniformity. A mapping rate below 90% suggests either contamination, adapter issues, or a reference that does not match the sample well. For ONT data, the same alignment principles apply, but you should also check for base quality issues that might affect SNP calling. Ultra-long reads may have lower quality at the ends, so consider whether trimming is needed before alignment. The [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible tutorials on quality control and alignment for long-read data. These tutorials walk through the specific tools and parameters used in standard long-read analysis pipelines. ### Variant Calling Before Phasing You need three types of variant calls before you can phase SVs: small variants (SNVs and indels), structural variants, and optionally tandem repeat variants. These can be called separately or together depending on your tools. Small variant calls provide the SNP anchors for phasing. For HiFi data, small variant calling is highly accurate, with precision and recall above 99.9% for SNVs. For ONT data, small variant calling is less accurate, so you may want to use a separate short-read dataset for SNP calling if one is available. Structural variant calls come from long-read SV callers. The specific caller matters less than the quality of the calls. You should filter SV calls based on read support, size, and quality scores before phasing. Low-confidence SV calls will produce unreliable phasing results. Tandem repeat variants are increasingly recognized as important for phenotypic variation, and some phasing tools now handle them. HiPhase jointly phases SNVs, indels, structural variants, and tandem repeat variants, which is useful when repeat expansions are relevant to your research question. ### Reference Genome Considerations The reference genome you align to affects both SV calling and phasing. Standard references have gaps and errors, particularly in repetitive regions. The telomere-to-telomere (T2T) assembly has closed many of these gaps, providing a complete reference that improves SV detection in previously intractable regions. For phasing, the reference affects the coordinates of your variants and the ability to align reads across complex regions. If you are working in a region that is poorly represented in the standard reference, consider whether a T2T reference or a species-specific reference would improve your results. ## Workflow 1: Joint Phasing with HiPhase ### When to Use This Workflow HiPhase is designed for PacBio HiFi data and jointly phases small variants, structural variants, and tandem repeat variants. This workflow is appropriate when you have HiFi reads, have called all variant types, and want a single phasing output that includes allele assignments for every variant. The tool uses a dual mode allele assignment approach for detecting large variants and applies the A* algorithm to phasing. It also includes logic that allows phase blocks to span breaks caused by alignment issues around reference gaps and homozygous deletions. This means the tool can maintain phasing continuity across regions that would break other phasing tools. ### Input Requirements You need the following inputs for HiPhase: 1. A BAM file of HiFi reads aligned to the reference genome 2. A VCF file with small variant calls (SNVs and indels) 3. A VCF file with structural variant calls 4. Optionally, a VCF file with tandem repeat variant calls The small variant VCF must contain genotype information for heterozygous SNPs. The structural variant VCF should contain high-confidence SV calls with read support information. ### Running the Phasing HiPhase is available as source code and a pre-compiled Linux binary with a user guide. The tool includes innate multi-threading, statistics gathering, and concurrent phased alignment output generation. The basic command structure is: ``` hiphase -r reference.fasta -b reads.bam -s small_variants.vcf -v sv_calls.vcf -o output_prefix ``` The tool produces a phased VCF file where each variant has a haplotype assignment. It also produces statistics about phase block sizes and switchflip errors. ### Interpreting the Output The phased VCF contains the same variants as your input VCFs, but with additional phasing information. Each variant is assigned to haplotype 1 or haplotype 2, or remains unphased if there was insufficient information. Check the phase block statistics to understand the quality of the phasing. A high phase block NG50 indicates that most of the genome was phased into long contiguous blocks. A high number of switchflip errors indicates that the phasing is unreliable in some regions. For each SV of interest, verify that it is within a phase block and that it has at least one heterozygous SNP anchor on the same haplotype. If the SV is at the edge of a phase block, the assignment may be less reliable. ## Workflow 2: Pre-Phasing with SVUPP ### When to Use This Workflow SVUPP takes a different approach. Instead of phasing all variant types together, it improves SV genotyping by incorporating read phasing information into genotype likelihoods. This workflow is appropriate when you have already called SVs and want to improve the accuracy of the genotype calls, particularly for SVs that are not near other SVs. The approach works with both ONT and PacBio HiFi data. It can take per-read phasing information from reference panel based phasing methods such as QUILT2 or from reference-free phasing methods such as WhatsHap. ### Input Requirements SVUPP requires: 1. SV calls from a caller such as cuteSV2 2. Per-read phasing information from WhatsHap or QUILT2 3. Long-read data (ONT or HiFi) The tool is written in Nextflow with a modular design, which means it can be integrated into existing Nextflow pipelines. The [nf-core documentation](https://nf-co.re/docs) provides guidance on running and configuring Nextflow pipelines, which is useful if you are building a custom workflow around SVUPP. ### Running the Workflow SVUPP is available at its GitHub repository. The Nextflow pipeline takes the SV calls and the per-read phasing information as inputs and produces genotyped SVs with phasing-aware likelihood scores. The key difference from standard SV genotyping is that SVUPP uses the phasing information to inform the genotype likelihood calculation. This means that reads supporting the SV are evaluated in the context of which haplotype they come from, instead of treating all reads equally. ### Interpreting the Output The output is a VCF file with SV genotypes that incorporate phasing information. The genotype likelihoods reflect whether the read support for the SV is consistent across both haplotypes or concentrated on one haplotype. In benchmarking, SVUPP achieved higher accuracy than cuteSV2, Sniffles2, and kanpig for genotyping SVs without close neighbor SVs, using both long and ultra-long ONT data as well as PacBio HiFi data. This improvement is most pronounced for SVs that are isolated from other SVs, where the phasing information provides additional signal for genotype assignment. ## Workflow 3: Reference-Panel Phasing ### When to Use This Workflow Reference-panel phasing uses population information to assign haplotypes. This approach is useful when you have a reference panel for your species or population, such as those available from [NCBI](https://www.ncbi.nlm.nih.gov/) for human populations. The advantage of reference-panel phasing is that it can phase variants that are not covered by any single read. The disadvantage is that it requires a reference panel that matches your sample population, and it may not capture rare or private variants. ### Input Requirements Reference-panel phasing requires: 1. SNP genotypes for your sample 2. A reference panel with phased haplotypes 3. A phasing tool that supports reference-panel phasing, such as QUILT2 The SNP genotypes can come from long-read data or from short-read data if available. The reference panel must be in a format compatible with the phasing tool. ### Running the Workflow The phasing tool uses the reference panel to infer haplotypes for your sample. The output is a set of phased SNP haplotypes that can be used as anchors for SV phasing. Once you have phased SNP haplotypes, you can assign SVs to haplotypes by identifying which haplotype carries the SNPs that are on the same reads as the SV. This step requires read-level information, so you need to examine the reads that support each SV call. ### Limitations Reference-panel phasing is limited by the quality and relevance of the reference panel. If your sample comes from a population that is not well represented in the panel, the phasing accuracy will be lower. Additionally, reference-panel phasing cannot phase variants that are not in the panel, so novel SVs will remain unphased unless they are connected to phased SNPs by reads. ## Workflow 4: Assembly-Based Phasing ### When to Use This Workflow Assembly-based phasing is the most computationally intensive approach but provides the most complete resolution of complex SVs. This workflow is appropriate when you are working with very large or complex structural variants that cannot be resolved by read-based phasing alone. The approach involves assembling the genome from long reads, producing haplotype-resolved contigs, and then identifying SVs by comparing the assembled haplotypes to the reference. This method was used to resolve ring chromosomes, Robertsonian translocations, and complex structural variants that were unresolved by short reads. ### Input Requirements Assembly-based phasing requires: 1. High coverage long-read data, ideally including ultra-long reads 2. An assembler that produces haplotype-resolved assemblies 3. A method for comparing the assembly to the reference genome For complex SVs, the combination of long-read sequencing and telomere-to-telomere assembly has enabled resolution of rearrangements in highly repetitive genomic regions that were previously intractable. ### Running the Workflow The assembly process produces contigs that represent individual haplotypes. By aligning these contigs to the reference genome, you can identify structural variants and determine which haplotype carries each variant. This approach was applied to 13 cases with ring chromosomes, Robertsonian translocations, and complex SVs that were unresolved by short reads. The analysis resolved 10 of 13 cases, including a Robertsonian translocation and all ring chromosomes. Multiple breakpoints were localized to genomic regions previously recalcitrant to sequencing, such as acrocentric p-arms, ribosomal DNA arrays, and telomeric repeats. ### Interpreting the Output The output of assembly-based phasing is a set of phased contigs with SV annotations. Each SV is assigned to a specific haplotype based on which contig carries it. This approach also enables additional analyses that are not possible with read-based phasing. For example, methylation profiling from long-read data can be used to discover phased differential methylation in gene promoters, revealing long-range position effects where a structural variant affects gene expression from a distance. ## Comparing Phasing Strategies ### Accuracy Considerations The accuracy of SV phasing depends on the method and the data. Joint phasing with HiPhase provides high accuracy for HiFi data because the reads are both long and accurate. The tool reported an average phase block NG50 of 480 kilobases with 929 switchflip errors, and fully phased 93.8% of genes in its assessment. Pre-phasing with SVUPP improves genotyping accuracy for SVs without close neighbor SVs. The improvement is achieved by incorporating read phasing information into genotype likelihoods, which reduces errors in genotype assignment. Assembly-based phasing provides the highest resolution for complex SVs but requires more data and computation. The success rate for resolving complex cases was 10 of 13 in the study of ring chromosomes and Robertsonian translocations. ### Cost and Time Considerations Read-based phasing with HiPhase or SVUPP is computationally efficient and can be run on a standard server. The main cost is the sequencing itself, particularly if you need high coverage for accurate phasing. Assembly-based phasing requires more computational resources and time. The assembly process is computationally intensive, and the downstream analysis of comparing haplotypes to the reference adds additional steps. ### Data Requirements HiFi data provides the best combination of read length and accuracy for phasing. The 99.8% accuracy and 13.5 kilobase average read length mean that a single read can provide reliable evidence for both the SV and the flanking SNPs. ONT data provides longer reads but lower accuracy. Ultra-long reads can span very large SVs and provide haplotype continuity across large regions, but the lower base accuracy means that SNP detection within a single read is less reliable. For the most complete phasing, combining HiFi and ultra-long ONT data can provide both accuracy and length. The HiFi data provides accurate SNP calls, while the ultra-long data provides haplotype continuity across very large regions. ## Practical Implementation Steps ### Step 1: Assess Your Data Before starting the phasing workflow, assess the quality and coverage of your long-read data. Check the read length distribution, the coverage depth, and the alignment statistics. For phasing, you need sufficient coverage to detect heterozygous SNPs and to have multiple reads supporting each SV. For HiFi data, aim for at least 20x coverage for variant calling and phasing. For ONT data, higher coverage may be needed to compensate for lower per-base accuracy. ### Step 2: Call Small Variants Call small variants from your long-read data. For HiFi data, use a variant caller designed for accurate long reads. For ONT data, consider whether you need a separate short-read dataset for accurate SNP calling. Filter the small variant calls to retain high-confidence heterozygous SNPs. These will serve as the anchors for phasing. ### Step 3: Call Structural Variants Call structural variants from your long-read data using an SV caller appropriate for your platform. Filter the SV calls based on read support and quality scores. If you are using the SVUPP workflow, you will need SV calls from a caller such as cuteSV2. If you are using HiPhase, you can use any SV caller that produces a VCF with the required fields. ### Step 4: Choose Your Phasing Strategy Based on your data and research question, choose one of the phasing strategies described above. For most applications with HiFi data, joint phasing with HiPhase is the most direct approach. For improving SV genotyping accuracy, SVUPP is appropriate. For complex SVs in repetitive regions, assembly-based phasing may be necessary. ### Step 5: Run the Phasing Run the phasing tool with your input data. Follow the tool-specific instructions for parameters and input formats. For HiPhase, ensure that your input VCFs are properly formatted and that the reference genome matches the one used for alignment. For SVUPP, ensure that the per-read phasing information is correctly generated from WhatsHap or QUILT2. ### Step 6: Evaluate Phasing Quality After running the phasing, evaluate the quality of the output. Check the phase block statistics, the number of switchflip errors, and the proportion of variants that were successfully phased. For each SV of interest, verify that it is within a phase block and that it has sufficient SNP anchors for confident assignment. ### Step 7: Validate with Independent Methods If possible, validate the phasing results with an independent method. Optical genome mapping was used to validate complex SV resolution in the ring chromosome study. For standard SVs, PCR-based validation or short-read sequencing can provide confirmation. ## Records and Measurements for Phasing Quality ### Phase Block Statistics Record the phase block NG50, which measures the length of the phased blocks. A higher NG50 indicates more contiguous phasing. Also record the number of phase blocks and the distribution of block sizes. ### Switchflip Error Rate Record the number of switchflip errors, which indicate where the phasing switches from one haplotype to the other within a block. A high switchflip error rate indicates unreliable phasing. ### Proportion of Variants Phased Record the proportion of SVs that were successfully phased. This proportion depends on the density of heterozygous SNPs near the SVs and the read length. ### Read Support for Phased SVs For each phased SV, record the number of reads that support the SV and the number of reads that connect the SV to a specific haplotype. Higher read support indicates more confident phasing. ## Common Failure Patterns in SV Phasing ### Insufficient Heterozygous SNP Density The most common cause of phasing failure is insufficient heterozygous SNP density near the SV. If the region around the SV is homozygous for all SNPs, there is no information to distinguish the two haplotypes. **Diagnosis**: Check the SNP density in the region flanking the SV. If there are no heterozygous SNPs within read length distance of the SV, phasing will fail. **Mitigation**: Use longer reads to connect the SV to more distant SNPs. Alternatively, use a reference panel to phase the region if one is available. ### Read Length Too Short for the SV If the SV is larger than the read length, no single read will span the entire variant. This makes phasing more difficult, though not impossible if reads cover the breakpoints and connect to SNPs on either side. **Diagnosis**: Compare the SV size to the read length distribution. If the SV is larger than the N50 read length, phasing will be challenging. **Mitigation**: Use ultra-long reads for very large SVs. The ability to generate reads exceeding 100 kilobases enables phasing of large deletions and inversions. ### Alignment Errors in Repetitive Regions Repetitive regions cause alignment errors that produce false SV calls and incorrect SNP genotypes. These errors corrupt phasing. **Diagnosis**: Check for SVs that are called in known repetitive regions. These may be alignment artifacts instead of true variants. **Mitigation**: Use a telomere-to-telomere reference that has resolved repetitive regions. The T2T assembly has closed many gaps in the human reference genome, improving SV detection in previously intractable regions. ### Low Coverage Low coverage reduces the number of reads supporting each variant and the number of reads connecting SVs to SNP anchors. This reduces phasing confidence. **Diagnosis**: Check the coverage in the region of interest. Low coverage regions will have fewer reads for phasing. **Mitigation**: Increase sequencing coverage or use a targeted approach to sequence the region of interest at higher depth. ### Platform-Specific Artifacts Each sequencing platform has specific error profiles that can affect phasing. ONT data has lower per-base accuracy, which can produce false SNP calls. HiFi data has higher accuracy but may have specific error patterns in homopolymer regions. **Diagnosis**: Compare variant calls from different platforms if you have multiple datasets. Discordant calls may indicate platform-specific artifacts. **Mitigation**: Use platform-appropriate variant callers and filters. For ONT data, consider using a separate short-read dataset for SNP calling. ## Limitations of SV Phasing ### Phasing Does Not Determine Parental Origin Phasing assigns variants to haplotypes, but it does not tell you which haplotype came from which parent. To determine parental origin, you need either parental genotypes or pedigree information. If you have parental data, you can compare the phased haplotypes to the parental genotypes to determine which haplotype was inherited from each parent. This information is important for understanding inheritance patterns and for clinical applications. ### Phase Blocks Have Boundaries Phasing does not produce a complete haplotype for the entire genome. Phase blocks are contiguous regions where variants have been assigned to haplotypes, but the boundaries between blocks occur where there is insufficient information to connect the haplotypes. If an SV of interest is near a phase block boundary, the phasing may be less reliable. You may need additional data or a different phasing strategy to resolve the region. ### Complex SVs May Not Be Fully Resolved Some SVs are too complex to be resolved by read-based phasing alone. Ring chromosomes, Robertsonian translocations, and complex rearrangements involving multiple breakpoints may require assembly-based approaches. The resolution of these complex SVs has been enabled by the combination of long-read sequencing, telomere-to-telomere assembly, and tools designed to discover chromosomal rearrangements from long reads. Even with these approaches, some cases remain unresolved. ### Reference Bias Affects Phasing Phasing is performed relative to a reference genome. If the reference does not match the sample well, alignment errors can produce false variants and incorrect phasing. The choice of reference genome matters. Standard references have gaps and errors, particularly in repetitive regions. The telomere-to-telomere assembly provides a more complete reference that improves SV detection and phasing in previously intractable regions. ## Safety and Reproducibility Context ### Reproducible Workflows Phasing analyses should be reproducible. This means documenting the exact versions of all tools, the parameters used, and the input data. Containerized workflows and pipeline managers help ensure reproducibility. The [nf-core documentation](https://nf-co.re/docs) describes community standards for pipeline development, including version control, containerization, and testing. Following these standards ensures that your phasing workflow can be rerun and shared with other researchers. The [Bioconductor project](https://bioconductor.org/) provides R packages for genomic analysis that can be used for downstream analysis of phased variants. These packages follow reproducible analysis standards and provide documentation for their use. ### Training and Skill Development Phasing structural variants requires skills in command-line computing, sequence alignment, variant calling, and genomic data analysis. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) program provides learning pathways for bioinformatics that cover these topics. The [Galaxy Training Network](https://training.galaxyproject.org/) offers accessible tutorials on long-read analysis and variant calling. These tutorials provide hands-on experience with the tools used in phasing workflows. For foundational computing skills, [The Carpentries](https://carpentries.org/lessons) lessons cover shell scripting, data organization, and reproducible computing practices that support complex genomic analyses. ### Data Management Phasing analyses generate large intermediate files, including aligned reads, variant calls, and phased output. Proper data management is essential for reproducibility and for avoiding data loss. Store raw sequencing data in a secure location with backup. Document the processing steps and the versions of all tools used. Use consistent file naming conventions and directory structures. The [NCBI](https://www.ncbi.nlm.nih.gov/) provides data repositories for sequence data and variant calls. Depositing your data in these repositories ensures that it is available for future analysis and for other researchers. ## Professional Escalation Criteria ### When to Seek Additional Expertise Some phasing problems require specialized expertise. Consider seeking help from a bioinformatics core facility, a collaborator with phasing experience, or a community support forum when: 1. Your phasing results are inconsistent across different tools or datasets 2. You are working with complex SVs that are not resolved by standard approaches 3. You need to phase variants in highly repetitive regions 4. You are uncertain about the interpretation of phasing results for clinical or diagnostic applications ### When to Consider Additional Sequencing If your phasing results are limited by data quality or coverage, consider whether additional sequencing would help. Ultra-long reads can resolve large SVs that are not spanned by standard long reads. Higher coverage can improve SNP detection and phasing confidence. ### When to Use Assembly-Based Approaches If read-based phasing fails to resolve SVs of interest, consider assembly-based approaches. These approaches are more computationally intensive but can resolve complex SVs that are intractable with read-based methods. The combination of long-read sequencing and telomere-to-telomere assembly has enabled resolution of ring chromosomes, Robertsonian translocations, and complex structural variants that were previously unresolved. If your SVs of interest fall into these categories, assembly-based phasing may be necessary. ## Frequently Asked Questions ### What is the difference between phasing and genotyping a structural variant? Genotyping determines whether a variant is homozygous or heterozygous, meaning whether both chromosomes carry the variant or only one. Phasing goes further and determines which specific chromosome carries the variant. For a heterozygous SV, genotyping tells you that one chromosome has the variant and the other does not, but it does not tell you which one. Phasing assigns the variant to a specific haplotype, which can then be traced to a parental chromosome if parental data is available. ### Can I phase structural variants with short-read sequencing data? Short reads are generally too short to phase structural variants directly because they cannot span both the SV and the flanking SNPs that anchor it to a haplotype. However, short reads can be used to call SNPs that are then phased with a reference panel, and the phased SNPs can be used as anchors for SV phasing if long reads connect the SV to those SNPs. This approach is less direct and less reliable than phasing with long reads. ### How much coverage do I need for SV phasing with HiFi reads? The coverage needed depends on the complexity of the region and the density of heterozygous SNPs. In general, higher coverage provides more reads supporting each variant and more reads connecting SVs to SNP anchors. For the human genome, coverage of 20x or higher with HiFi reads has been shown to provide accurate variant detection and phasing. Lower coverage may be sufficient for simple SVs in SNP-dense regions, while higher coverage may be needed for complex regions. ### What is a phase block and why does it matter for SV phasing? A phase block is a contiguous genomic region where variants have been assigned to one of two haplotypes. The boundaries between blocks occur where there is insufficient information to connect the haplotypes across the gap. For SV phasing, the relevant question is whether the SV falls within a phase block that contains enough SNP anchors to assign it confidently. If the SV is near a block boundary, the assignment may be less reliable. ### How do I determine which parent contributed a phased SV? Phasing assigns variants to haplotypes, but it does not determine parental origin. To determine which parent contributed a specific SV, you need either parental genotypes or pedigree information. If you have parental data, you can compare the phased haplotypes to the parental genotypes to determine which haplotype was inherited from each parent. This information is important for understanding inheritance patterns and for clinical applications. ### What tools are available for phasing structural variants with long reads? Several tools are available for phasing structural variants with long reads. HiPhase jointly phases SNVs, indels, structural variants, and tandem repeat variants from HiFi sequencing. SVUPP improves SV genotyping by incorporating read phasing information into genotype likelihoods and can be applied together with SV callers such as cuteSV2. WhatsHap provides reference-free phasing, and QUILT2 provides reference-panel based phasing. The choice of tool depends on your data type and research question. ### Can I phase SVs in repetitive regions of the genome? Repetitive regions are challenging for SV phasing because alignment errors are more common and because standard references have gaps in these regions. The telomere-to-telomere assembly has closed many of these gaps, improving SV detection and phasing in previously intractable regions. Ultra-long reads can also help by spanning large repetitive regions. For the most complex cases, assembly-based phasing may be necessary. ### What should I do if my SV of interest is not phased? If an SV of interest is not phased, first check whether there are heterozygous SNPs near the SV. If the region is homozygous for all SNPs, there is no information to distinguish the two haplotypes, and the SV cannot be phased regardless of read length. If there are heterozygous SNPs but the SV is still not phased, the issue may be read length, coverage, or alignment quality. Consider whether additional sequencing, longer reads, or a different phasing strategy would help. ## Related Bioinformatics Guides - [Detecting Structural Variants with Long-Read Sequencing: Methods and Considerations](/knowledge/bioinformatics/detecting-structural-variants-with-long-read-sequencing-methods-and-considerations) - [De Novo Genome Assembly with Long Reads: A Practical Workflow](/knowledge/bioinformatics/de-novo-genome-assembly-with-long-reads-a-practical-workflow) - [RNA Sequencing Data Analysis: From Raw Reads to Differential Expression](/knowledge/bioinformatics/rna-sequencing-data-analysis-from-raw-reads-to-differential-expression) - [Cell Cycle Checkpoints: A Decision Framework for Identifying Phase-Specific Defects](/knowledge/bioinformatics/cell-cycle-checkpoints-a-decision-framework-for-identifying-phase-specific-defects) - [RNA-Seq vs Microarray: Choosing the Right Gene Expression Profiling Platform](/knowledge/bioinformatics/rna-seq-vs-microarray-choosing-the-right-gene-expression-profiling-platform) ## 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. - [SVUPP: Pre-phasing long reads improves structural variant genotyping.](https://pubmed.ncbi.nlm.nih.gov/41134129). Bioinformatics (Oxford, England), 2022. - [Accurate circular consensus long-read sequencing improves variant detection and assembly of a human genome.](https://pubmed.ncbi.nlm.nih.gov/31406327). Nature biotechnology, 2019. - [Nanopore sequencing and assembly of a human genome with ultra-long reads.](https://pubmed.ncbi.nlm.nih.gov/29431738). Nature biotechnology, 2018. - [Resolution of ring chromosomes, Robertsonian translocations, and complex structural variants from long-read sequencing and telomere-to-telomere assembly.](https://pubmed.ncbi.nlm.nih.gov/39520989). American journal of human genetics, 2024. - [HiPhase: jointly phasing small, structural, and tandem repeat variants from HiFi sequencing.](https://pubmed.ncbi.nlm.nih.gov/38269623). Bioinformatics (Oxford, England), 2024. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.