Haplotype Phasing Long Reads
Haplotype phasing with long reads assigns sequence variants to specific parental chromosomes, producing haplotype resolved genomes. This guide is for bioinformaticians, genome biologists, and clinicians who need to phase diploid genomes using long read sequencing data. It applies to PacBio HiFi, Oxford Nanopore, and other long read platforms. For a detailed workflow introduction, see the Galaxy Training Network.
Phasing is essential for understanding compound heterozygosity, allele specific expression, and structural variant inheritance. Long reads span multiple heterozygous sites in a single fragment, enabling direct read based phasing without requiring pedigree data. This approach contrasts with population based phasing methods that rely on linkage disequilibrium. The EMBL-EBI Training provides foundational lessons on these concepts.
At a Glance
| Concept | Key Point | Reference |
|---|---|---|
| Haplotype | Set of alleles inherited from one parent on a contiguous chromosome segment | NCBI Bookshelf |
| Phasing | Process of determining which homologous chromosome carries which allele | Haplotype-aware long-read error correction |
| Long read phasing | Directly phases variants using reads that span multiple heterozygous sites | Comparison of nanopore sequencing... |
| Phase block | Contiguous region where phasing is consistent and not broken | LongAllele: a joint inference framework |
| Switch error | Incorrect orientation of alleles across adjacent phase blocks | Bioconductor |
| Coverage requirement | Minimum 30x to achieve high quality phasing with long reads | Telomere-to-telomere assembly... |
Core Concepts
A haplotype is a combination of alleles at multiple loci on a single chromosome. Phasing resolves which alleles are inherited together. Long read sequencing produces fragments that often contain two or more heterozygous variant sites, allowing direct inference of the haplotype phase without statistical imputation. The NCBI Bookshelf describes the genetic basis of haplotypes and recombination.
Long read phasing relies on reads that cover two or more heterozygous single nucleotide variants or small indels. Each read provides a partial haplotype. Overlapping reads are assembled into larger phase blocks. Tools such as WhatsHap, LongPhase, and MarginPhase implement this approach. The Haplotype-aware long-read error correction paper describes how error correction influences phasing accuracy.
Key terms include phase block N50 (the block length at which 50% of phased bases are in blocks of that size or longer) and switch error rate (the proportion of positions where the assigned phase flips relative to the true phase). Phase blocks can be broken by low coverage regions, repetitive sequences, or strong mapping biases.
Decision Criteria
Selecting a phasing method depends on read length, sequencing error rate, coverage depth, and genomic heterozygosity. Long reads with low error rates, such as PacBio HiFi (Q20+), produce higher phase block N50 and fewer switch errors. Oxford Nanopore reads, especially after correction, can also phase effectively but may require deeper coverage. The Comparison of nanopore sequencing... study outlines trade offs.
Coverage of at least 30x is recommended for robust phasing. Higher heterozygosity improves phasing because more informative sites are present. For genomes with low heterozygosity (e.g., inbred organisms), phasing may require pedigree data or additional sequencing. The Proband Nanopore Long-Read Genome Sequencing... paper demonstrates phasing in a clinical context.
Computational resources matter. Phasing large genomes requires memory proportional to read depth and block size. Use of vectorized algorithms or distributed computing may be necessary. The Galaxy Training Network offers scalable workflow options.
Practical Workflow or Implementation Steps
Data acquisition and quality control. Obtain long read data from public repositories such as the NCBI Sequence Read Archive. Run base calling (for raw signal data), adapter trimming, and quality filtering. Evaluate read length distribution and per base quality.
Read alignment. Map reads to a reference genome using a long read aware aligner such as minimap2 or Winnowmap. Sorting and indexing are required. The Bioconductor ecosystem provides functions for BAM file manipulation.
Variant calling. Call small variants (SNVs and indels) from the aligned reads. Use tools like DeepVariant, Clair3, or PEPPER. Filter calls by quality and depth. Retain only heterozygous variants for phasing.
Phasing. Run a long read phasing tool on the VCF file and the read alignments. For example, WhatsHap phase requires a BAM file and a VCF. The output is a phased VCF with phase blocks. The Telomere-to-telomere assembly using HERRO-corrected Nanopore Simplex reads describes how phasing contributes to complete genome assembly.
Phasing validation. Compare phased blocks against known haplotypes from pedigree or population data. Calculate switch error rate and phase block N50. Use tools like whatshap stats or custom scripts.
Visualization and integration. Visualize phase blocks using IGV or similar viewers. Integrate with downstream analyses such as allele specific expression (see LongAllele: a joint inference framework).
Quality Checks
Verify that alignment depth is uniform across heterozygous sites. Low coverage regions can break phase blocks. Use samtools depth to inspect coverage. The EMBL-EBI Training provides training on alignment QC.
Calculate the switch error rate if a truth set is available. For human genomes, reference trios or the Genome in a Bottle samples can serve as validation. Without a truth set, use phasing concordance within large blocks. High quality phasing typically shows switch error rates below 1% for HiFi data.
Check phase block N50. For a human genome with 30x HiFi, expect N50 above 1 megabase. Lower values may indicate insufficient coverage, high error rates, or low heterozygosity. The NCBI Bookshelf resource offers metrics descriptions.
Inspect for systematic mapping bias, especially in low complexity regions. Phasing in repeats or segmental duplications may be unreliable. The Annotated, chromosome-level genome for the spotted turtle illustrates handling of repetitive content.
Common Mistakes
- Skipping read error correction. High error rates disrupt variant calling and phasing. Correcting reads with tools like HERRO or Canu improves phasing. See Haplotype-aware long-read error correction.
- Insufficient coverage. Coverage below 20x yields fragmented phase blocks and high switch error rates. Always verify coverage after mapping.
- Ignoring read alignment quality. Soft clipping, chimeric reads, and poor mapping quality degrade phasing. Filter reads with low mapping quality.
- Using only one phasing tool. Different tools have strengths. WhatsHap is fast but may be sensitive to errors. LongPhase handles structural variants. Compare outputs.
- Failure to remove duplicates. PCR duplicates or optical duplicates create false haplotype support. Mark duplicates before phasing.
Limits of Interpretation
Phasing is only accurate within phase blocks. Blocks can break due to recombination, repetitive DNA, or sequencing gaps. The boundaries between blocks are uncertain unless validated by independent methods (e.g., trio data or strand sequencing). The Comparison of nanopore sequencing... study notes that methylation phased reads can provide additional confirmation.
Phasing is impossible in regions where no heterozygous sites exist. In highly inbred organisms or homozygous genomes, long read phasing yields few to no blocks. Complementary approaches such as linked read sequencing or optical mapping may be necessary.
Long read phasing does not resolve complex structural variants on its own. Phase blocks spanning large insertions or deletions may be incorrect if the variant is not represented in the reference. The Telomere-to-telomere assembly work shows the need for assembly based phasing in such cases.
Switch errors can occur at any block boundary. Switch error rates increase with decreasing read length and coverage. Do not interpret phase across blocks as confirmed. Use statistical tests or parental data to verify long range phasing.
Frequently Asked Questions
What is the minimum coverage for good phasing with long reads?
For high accuracy phasing with PacBio HiFi, 30x coverage is recommended. For Oxford Nanopore data, 40x to 60x may be needed depending on error rate. Lower coverage can work for targeted regions but yields fragmented phase blocks.
Can I phase with short reads instead of long reads?
Short read phasing relies on population linkage information or statistical inference. Long reads provide direct molecular phasing that is more accurate for individual genomes. Short read phasing often produces shorter phase blocks and higher switch error rates.
How do I assess phasing accuracy without a truth set?
Calculate phase block N50 and the distribution of block lengths. High N50 with few breaks suggests good phasing. You can also check consistency of phasing across overlapping reads using tools like whatshap stats. Comparison with a second independent phasing run using a different tool can reveal discrepancies.
Why is read error correction important for phasing?
High error rates cause mismapping and false variant calls. Error correction reduces false positive heterozygous sites and improves alignment, leading to longer, more accurate phase blocks. The Haplotype-aware long-read error correction publication demonstrates this effect.
References and Further Reading
- NCBI Bookshelf , Comprehensive biomedical textbooks covering haplotype biology.
- EMBL-EBI Training , Interactive courses on phasing and long read analysis.
- Galaxy Training Network , Workflow tutorials for long read phasing.
- Bioconductor , R packages for phasing assessment and visualization.
- NCBI Sequence Read Archive , Repository for long read sequencing data.
- Haplotype-aware long-read error correction , Algorithms for error correction improving phasing.
- Comparison of nanopore sequencing, MethylationEPIC array, and EM-Seq for DNA methylation detection , Context on phasing and methylation.
- An annotated, chromosome-level genome for the spotted turtle, Clemmys guttata , Phasing in a non model organism.
- Proband Nanopore Long-Read Genome Sequencing Facilitates Preimplantation Genetic Testing for Facioscapulohumeral Muscular Dystrophy , Clinical application of phasing.
- LongAllele: a joint inference framework for allele-specific analysis on long-read bulk and single-cell RNA sequencing , Phasing for allele specific expression.
- Telomere-to-telomere assembly using HERRO-corrected Nanopore Simplex reads , High quality phasing in complete genomes.