Long-Read Genome Assembly: Planning a Reproducible Project

If you are planning a de novo genome assembly using Oxford Nanopore or PacBio long reads, this guide is for you. It covers the essential steps from sample quality assessment through to a reproducible, versioned final assembly. You will learn how to set coverage targets, choose an assembler, polish base calls, validate contiguity and accuracy, and document every parameter. The goal is not a single "best" pipeline but a framework you can adapt to your organism and data type.

At a Glance

Sample Quality and Input DNA

The assembly starts in the lab. For long reads, high molecular weight DNA is critical. Degraded DNA shortens read lengths and lowers continuity. Use a Qubit fluorometer for accurate quantitation and a TapeStation or Femto Pulse to confirm fragment sizes above 20 kb for ONT, or 10 kb for PacBio HiFi. The Galaxy Training Network (source 3) offers hands on tutorial material for checking input quality. For extremely large or GC rich genomes, consider multiple extraction methods. A single poor sample can waste weeks of compute.

Coverage Planning

Coverage is the depth of sequence needed. For a haploid genome of size G, plan for total bases = coverage depth x G. Long read assemblers typically perform well at 30x to 60x for a moderately heterozygous diploid. For genomes with high repeats or polyploidy, aim for 60x to 100x. Heterozygosity inflates the effective genome size, use a kmer survey with tools like GenomeScope before sequencing. The NCBI Sequence Read Archive (source 5) hosts raw data from many published assemblies that can inform your coverage decisions. For example, the ONT only assembly of a Korean male genome in source 11 (Genes Genomics) used 60x coverage to achieve chromosome scale contiguity. Plan extra capacity for polishing and validation.

Choosing an Assembler

No single assembler is optimal for all data. Flye works well for ONT and PacBio CLR, Canu is robust for noisy reads, NextDenovo is fast for large genomes, Shasta suits ONT only. For PacBio HiFi, try HiCanu or hifiasm. The EMBL EBI Training (source 2) provides workflows that compare assemblers on real datasets. Key decision criteria include:

• Read type and error profile.
• Genome size and complexity.
• Available memory and CPU.
• Desired contiguity vs. compute time.

For small bacterial genomes, canu or flye are adequate. For mammalian genomes, hifiasm (HiFi) or NextDenovo (ONT) often produce the most contiguous results. Always test with a subset before running full assembly. The de novo assembly of Heliconius butterflies in source 6 used Flye and demonstrates that even with long read only data, high contiguity is achievable.

Polishing

After the initial assembly, base accuracy is typically 85-98%. Polishing corrects errors using the raw signal (ONT) or aligned reads. For ONT, Medaka or Racon are common, for PacBio, Arrow or Merfin. Run at least two rounds. Over polishing can introduce false substitutions, validate with kmer counts. The Bioconductor (source 4) ecosystem includes packages like R for assembly evaluation. The grapevine genome methylome map in source 7 used iterative polishing to resolve methylation patterns, highlighting the importance of careful parameter tuning.

Validation

A good assembly is contiguous, complete, and free of structural misassemblies. Use metrics:

• N50 and L50 for contiguity.
• BUSCO completeness score against a lineage dataset.
• Merqury for kmer based consensus accuracy.
• Assemblytics or dot plots to detect misjoins.

The NCBI Bookshelf (source 1) contains a chapter on genome assembly validation. For the rat tapeworm in source 9, BUSCO completeness reached 97% only after alignment based validation. Compare to closely related references if available. Validate with short reads or optical maps if budget allows.

Versioned Reporting

Reproducibility requires tracking every parameter. Use a workflow manager like Snakemake or Nextflow, and store all commands, versions, and intermediate files. Write a YAML config file with genome size, coverage, assembler version, and polishing rounds. Commit to Git. The Galaxy Training Network (source 3) also provides shared histories and workflows that can be version controlled. Report the final assembly as a FASTA, plus statistics, BUSCO results, and a methods paragraph that would allow a colleague to replicate the process.

Common Mistakes

• Insufficient coverage: 15x or less rarely yields a contiguous assembly.
• Skipping quality control on reads: adapter contamination, low quality tails.
• Using default parameters without checking: especially for very large or small genomes.
• Over polishing: up to 3 rounds, then stop.
• Not validating with an independent method (short reads, optical map, or genetic map).
• Forgetting to submit raw data to SRA or ENA, which hinders reproducibility.

Limits and Uncertainty

Long read assemblies still struggle with centromeres, telomeres, and very high repeat regions. Telomere to telomere assemblies are currently only possible with combined ultra long reads and manual curation. Heterozygous regions may collapse or remain in two haplotigs. The dual genome map of grapevine in source 7 demonstrates that even with long reads, structural variants can be misassembled. Always treat the assembly as a first draft, later improvement with long range data (Hi C, optical maps) is common.

Frequently Asked Questions

1. What is the minimum coverage for a bacterial genome assembly? For a 5 Mb genome with ONT reads, 30x is often enough for a single contig. For PacBio HiFi, 20x may suffice. Always perform a subset test.

2. Should I use short reads to polish a long read assembly? It can help, but be careful. Short reads may fix errors but can also collapse repeats. Use only if you have high quality short reads and validate with kmer methods.

3. How do I handle polyploid genomes? Use assemblers designed for polyploids like hifiasm phased with HiFi, or use a deduplication tool. Expect more fragmented contigs. The Trypanosoma assembly in source 10 is a good chromosome level example for a haploid parasite.

4. What if my assembly has many small contigs? Check coverage, low depth may cause breaks. Try a different assembler or increase coverage. Also look for contamination or high repeat content.

References and Further Reading

• NCBI Bookshelf: Genome Assembly Validation (Source 1)
• EMBL EBI Training: Long Read Assembly Workflows (Source 2)
• Galaxy Training Network: Assembly Tutorials (Source 3)
• Bioconductor: Assembly Evaluation Packages (Source 4)
• NCBI Sequence Read Archive: Search for Long Read Datasets (Source 5)
• de novo assembly of 5 Heliconius butterfly genomes with long read sequencing alone (PubMed) (Source 6)
• A dual genome methylome map of clonal evolution in grapevine (PubMed) (Source 7)
• Complex structural variation in the mucin pangenome (medRxiv) (Source 8)
• Assembly of a high quality reference genome for the rat tapeworm (bioRxiv) (Source 9)
• Chromosome level gap free genome assembly of Trypanosoma larimichthysi (Sci Data) (Source 10)
• ONT only genome assembly of a Korean male individual (Genes Genomics) (Source 11)

Related Articles

• RNA Sequencing Analysis: From FASTQ Files to Biological Questions
• RNA-seq Quality Control: What to Check Before Differential Expression
• How to Plan a Bulk RNA-seq Differential Expression Study
• Single-Cell RNA-seq Workflow: A Practical Analysis Roadmap
• Single-Cell RNA-seq Quality Control: Cells, Genes, and Mitochondrial Reads