Short-read sequencing built the reference genomes that plant and animal researchers rely on every day. It also left gaping holes where repeats, segmental duplications, and centromeric regions sit — features that happen to be disproportionately important in crop domestication traits and livestock adaptation. Two technologies have changed that calculation. PacBio HiFi reads routinely exceed 15 kb with consensus accuracy above 99.9%, and Oxford Nanopore reads can stretch past 100 kb on a single molecule. Together they have made chromosome-level, and in many cases telomere-to-telomere, assemblies achievable for species where a finished genome seemed unreachable five years ago.

This article walks through what each platform actually delivers for plant and animal genome assembly — not the marketing numbers, but the assembly outcomes researchers see with real genomes. It covers where the two platforms diverge, what polyploidy and repeat-rich livestock genomes demand from an assembly strategy, which assemblers earn their place in a production pipeline, and what to look for when you decide to outsource rather than build in-house.

Why Short Reads Hit a Wall

A 150-bp paired-end read cannot span a 5-kb tandem repeat, much less a 50-kb segmental duplication. The assembler sees identical sequences at multiple positions and collapses them into a single representation — a compression artifact that erases paralogs, merges haplotypes, and discards structural variants. For crops with recent whole-genome duplications (wheat, sugarcane, Brassica species) and livestock with large repetitive families (cattle immune gene clusters, avian microchromosomes), this is not a minor inconvenience. It is the difference between a genome that maps functional variation and one that hides it.

Long reads solve the span problem directly. When a single read covers an entire repeat unit or a structural variant breakpoint, there is no ambiguity to resolve — the read itself provides the evidence for the correct layout. The assembly graph simplifies from a tangle of collapsed repeats to a sparse, well-supported structure. A 2022 review of complex genome assembly strategies documented this progression across dozens of species and concluded that the switch from short-read to long-read assembly consistently recovers 30-50% more contiguous sequence, with the largest gains in the most repetitive genomes [1].

The practical consequence for researchers is clear. If your species has a publicly available short-read assembly, it almost certainly underrepresents the gene families, regulatory elements, and structural variants that sit in repetitive regions. A long-read assembly is not an incremental improvement — it is a fundamentally different view of the genome. For researchers who also plan to run transcriptomic studies on the same species, the outsourcing transcriptome analysis guide covers how a high-quality reference genome improves RNA-seq alignment and differential expression accuracy in agricultural projects.

PacBio HiFi and Nanopore, Side by Side

The platforms share a common logic — sequence single molecules in real time — but diverge in chemistry, error profile, and practical tradeoffs that matter for assembly.

The numbers matter less than what they mean for assembly. HiFi reads are short enough (15–25 kb) that repetitive regions longer than the read still require bridging by assembly graph heuristics, but the per-base accuracy means the assembled contigs need little or no polishing. ONT reads are long enough to span most repeats in a single molecule, but the systematic errors — particularly homopolymer length inaccuracies in AT-rich regions — mean the consensus sequence usually requires a polishing step with short reads, HiFi reads, or a deep-learning corrector such as Medaka or HERRO.

Neither platform is universally better. The choice depends on the genome, not the technology.

Figure 1: Comparison of PacBio HiFi (circular consensus) and Oxford Nanopore (direct sequencing) workflows for de novo genome assembly, highlighting the tradeoffs between read accuracy and read length that drive platform selection for plant and animal genomes.

Where Each Platform Excels

HiFi reads produce the highest-quality assemblies for diploid and simple-polyploid genomes where a single read can resolve most heterozygous sites and repeat lengths stay under 20 kb. For maize (2.3 Gb, 85% repetitive), rice (430 Mb, compact), and most diploid livestock (cattle, pig, chicken), HiFi-only assemblies now reach chromosome-level contiguity with a single SMRT Cell on Revio. Sequencing platforms serving agricultural genomics increasingly default to HiFi for these species because the assembly workflow is mature and requires minimal manual curation.

ONT excels where read length is the binding constraint. Large segmental duplications (50–200 kb) in Brassica and legume genomes, centromeric satellite arrays in wheat and barley, and the highly duplicated immune gene clusters in cattle and pigs all demand reads longer than what HiFi can deliver. ONT also has an advantage for field-deployable projects — the MinION is portable, and library preparation is faster — though throughput for full-scale genome assembly still requires a PromethION flow cell. For transcript-level applications of nanopore technology, the direct RNA sequencing applications in plant transcriptomics guide covers how the same platform supports full-length isoform characterization and RNA modification detection without reverse transcription.

For the growing number of projects pursuing telomere-to-telomere (T2T) assemblies, the strongest results come from combining both platforms. HiFi provides the accurate backbone, ONT reads span the gaps, and the resulting hybrid assembly reaches completeness levels that neither technology can achieve alone. Long-read data also enable direct detection of DNA base modifications — 5mC from PacBio kinetics or 6mA from nanopore raw signal — creating a bridge to the epigenomic sequencing approaches for crop research that add functional annotation to the assembled sequence.

Polyploid Genomes Demand More

Plant genomes introduce a challenge that most animal genomes avoid: polyploidy. Wheat is hexaploid. Sugarcane is an auto-octoploid with aneuploid sectors. Strawberry is octoploid. Even diploid crops such as soybean and maize carry the scars of ancient whole-genome duplications that complicate haplotype phasing.

Figure 2: Polyploid genome assembly challenges in crops — homeologous chromosomes require subgenome-level phasing and parental k-mer data to avoid collapsing distinct subgenomes into a single representation.

Haplotype-resolved assembly — where each parental haplotype is reconstructed as a separate chromosome-scale sequence — has become the standard for polyploid crop genomes. A 2025 review of haplotype-resolved assembly in polyploid plants documented the shift from collapsed consensus assemblies to phased diploid representations and described the tools and sequencing depths that make this feasible for genomes up to octoploid complexity [2].

The practical takeaway for researchers planning a crop genome project is that polyploidy is not a reason to settle for a fragmented assembly. It is a reason to plan for higher sequencing coverage (40–60× HiFi for tetraploids, 60–80× for hexaploids) and to choose an assembler that explicitly models polyploid inheritance. More on that in the next section. For background on how plant genome projects are structured from study design through data delivery, the plant genome sequencing resource provides an overview of service types and workflow decisions.

Livestock Genomes Have Their Own Rules

Animal genomes are not simply "easier" than plant genomes — they present a different set of challenges. Cattle, sheep, and pig genomes carry large segmental duplications in immune gene clusters (MHC, KIR, immunoglobulin loci) that span 100–500 kb and differ between breeds. Avian genomes pack their gene density into microchromosomes that are GC-rich and difficult to amplify uniformly. Aquaculture species (salmonids, carp, shrimp) combine recent polyploidy with extremely high heterozygosity from outbred populations.

Third-generation sequencing has been particularly transformative for livestock. A 2024 review of third-generation sequencing in livestock documented how long reads resolved previously intractable regions — including the cattle Y chromosome, the pig MHC, and salmonid homeologous chromosomes — that short-read assemblies had either collapsed or left entirely unassembled [3]. The pattern across species is consistent: long reads do not just improve contiguity metrics. They recover functional genomic regions that short-read assemblies systematically lose.

One livestock-specific consideration is sample availability. For elite breeding animals (a prize bull, a high-index sow), the research sample is often a single individual — there is no population of biological replicates to average over. This makes read length and assembly accuracy non-negotiable because there is only one shot at getting the genome right. For aquaculture species, the challenge is the opposite: high heterozygosity from outbred populations means the assembler needs enough coverage to resolve two distinct haplotypes from a single individual, which demands 30–40× HiFi coverage or 50–60× ONT coverage even for modest-size genomes. The NGS technologies in animal and plant breeding article covers how sequencing approaches map to different breeding program goals.

Assemblers That Actually Work

Choosing an assembler is not about feature lists. It is about which tool produces a correct assembly for your genome class at the coverage depth you can afford. The table below summarizes assemblers that have been validated on plant and animal genomes in published benchmarks.

The hifiasm family has become the de facto standard for most plant and animal genome projects. The telomere-to-telomere extension of hifiasm, described in a 2024 Nature Methods publication, introduced an ultra-long-aware mode that uses ONT reads exceeding 100 kb to bridge the repetitive gaps that HiFi reads alone cannot resolve, achieving T2T contiguity for several human chromosomes and, more recently, for plant genomes [4].

A 2023 benchmark that evaluated five long-read assemblers on plant and animal genomes with known reference sequences found that no single assembler dominated across all metrics — hifiasm produced the most contiguous assemblies, Flye handled the most repetitive genomes efficiently, and hybrid approaches (Verkko, hifiasm-UL mode) were necessary for T2T completeness [5]. The takeaway is practical: test at least two assemblers on a subset of your data before committing to a full-scale assembly, because assembler performance is genome-specific. For projects that need assembly plus downstream analysis as a single package, agricultural long-read sequencing data analysis services combine assembly with annotation, variant calling, and comparative genomics.

Figure 3: End-to-end de novo genome assembly pipeline for long-read data, from raw read preprocessing through assembly graph construction, scaffolding, polishing, and quality assessment with BUSCO completeness and merqury k-mer spectra.

When Outsourcing Makes Sense

Building an in-house long-read assembly pipeline is a significant commitment. It requires high-molecular-weight DNA extraction expertise (tissue-specific protocols, not off-the-shelf kits), access to a PacBio Revio or ONT PromethION, sufficient compute infrastructure (512 GB–1 TB RAM for large plant genomes, GPU nodes for basecalling and polishing), and — most critically — someone who has assembled enough genomes to recognize when the output looks wrong.

Outsourcing makes sense when any of these conditions apply:

The genome is large (>5 Gb) or polyploid. These assemblies stress compute resources and require assembler tuning that benefits from experience across multiple species. A de novo genome sequencing provider with polyploid-specific pipelines has likely encountered and solved the homeolog-collapse and coverage-allocation problems that first-time assemblers spend months debugging.

You only need one or a few genomes. The per-sample cost of outsourcing is lower than the fixed cost of instrument access, consumables, and compute for a small number of projects.

The project needs both sequencing and analysis as a deliverable. Coordinating sequencing at one facility and analysis at another introduces handoff errors and version mismatches that integrated providers avoid.

You need the assembly for downstream work, not for methods development. If the genome is a means to an end (GWAS, QTL mapping, genome editing target identification), a professionally assembled and annotated genome accelerates your actual research question.

The alternative — running assembly in-house — makes sense for groups that will assemble dozens of genomes and can amortize instrument and compute costs across many projects. It is also appropriate for methods-development labs that need control over every parameter. For everyone else, the assembly outcome matters more than who clicks "run."

Questions to Ask Any Provider

Not all sequencing and assembly providers deliver the same product. Asking the right questions during project scoping prevents unpleasant surprises when the data arrive.

What coverage depth do you recommend for my species, and why? The answer should reference genome size, heterozygosity, repeat content, and ploidy — not a generic "30×." For a hexaploid wheat genome, 30× HiFi coverage means 5× per subgenome, which is inadequate for haplotype-resolved assembly.

Which assembler will you use, and what is your QC process? The provider should name specific assemblers (not "we evaluate several options") and describe their QC checks: contiguity metrics (N50, L50, number of contigs), completeness (BUSCO scores for the appropriate lineage dataset), and consistency (merqury k-mer plots, coverage distribution). Generic answers signal a black-box pipeline.

Will I receive the raw data, the intermediate assembly, and the analysis files? You want the subreads or FAST5/POD5 files, the contig-level and scaffold-level assemblies, the assembly graph if applicable, and the QC reports. Retaining raw data access matters because assembly algorithms improve over time — a genome you assemble today can be reassembled better in two years from the same raw reads.

What does the timeline look like for my genome size and complexity? Realistic timelines for a service provider are 4–8 weeks from sample receipt to draft assembly for a diploid genome under 3 Gb, and 8–16 weeks for large or polyploid genomes. Timelines shorter than this suggest an automated pipeline with minimal manual curation.

Can you handle my sample type and what are the DNA requirements? High-molecular-weight DNA extraction from plant tissue (especially secondary-metabolite-rich leaves, woody stems, or seeds) and from certain animal tissues (adipose, connective tissue) requires specialized protocols. Ask whether the provider has extracted HMW DNA from your specific tissue type before.

These questions serve a dual purpose: they surface the technical details that determine assembly quality, and they reveal whether the provider is running a thoughtful operation or a high-throughput factory. A provider that cannot answer them clearly should not be assembling your genome.

Frequently Asked Questions

Q: What is the minimum coverage needed for a de novo genome assembly with long reads?

A: For PacBio HiFi, 20–30× coverage is the current consensus minimum for a diploid genome with moderate heterozygosity, generating contig N50s in the megabase-to-chromosome range for genomes under 3 Gb. For polyploid genomes, coverage should scale approximately with ploidy level — 40–60× for tetraploids and 60–80× for hexaploids — because each subgenome needs its own adequate representation. For Oxford Nanopore, 40–60× coverage is typically recommended to compensate for lower raw accuracy, though ultra-long ONT reads (100 kb+) used in hybrid assembly with HiFi can be effective at lower ONT coverage (20–30×).

Q: Can I assemble a highly polyploid genome with long reads alone, or do I need other data?

A: Long reads can produce chromosome-level assemblies for genomes up to hexaploid complexity (wheat, oat) without ancillary data, provided coverage is sufficient and the assembler models polyploid inheritance. For auto-polyploids with high heterozygosity (sugarcane, alfalfa, potato), adding parental short-read data for trio-binning through hifiasm substantially improves haplotype phasing and reduces switch errors. For octoploid and higher levels (strawberry, some ornamental species), Hi-C data resolves chromosome-scale scaffolding that assemblers alone cannot determine from reads, and is recommended as a complement to long-read assembly rather than a replacement.

Q: How do I decide between PacBio HiFi and Oxford Nanopore for my species?

A: The decision should be driven by genome characteristics, not platform loyalty. If your genome is under 3 Gb, diploid or allotetraploid, and repeat content is moderate, HiFi alone will likely produce the best assembly at the lowest per-base cost. If your genome exceeds 5 Gb, contains large segmental duplications (50–200 kb), or you need to resolve centromeric and telomeric regions for a T2T assembly, ONT ultra-long reads combined with HiFi data produce more complete assemblies. If you are working with a species that has no existing reference and you need maximum contiguity on a first pass, a hybrid HiFi-plus-ONT strategy with an assembler such as Verkko or hifiasm(UL) gives you the best shot at chromosome-level completeness without iterative improvement cycles.

Q: What deliverables should I expect from a de novo genome assembly project?

A: A complete assembly deliverable should include raw sequencing data (subreads or POD5/FAST5 files), the contig-level assembly (FASTA), the scaffolded assembly if Hi-C or genetic maps were used, assembly QC reports (N50, L50, BUSCO completeness, merqury k-mer spectra, coverage plots), the assembly graph in GFA format for downstream curation and visualization, and any annotation files if included in the scope of work (gene models, repeat annotations, functional predictions). Always confirm before the project starts that raw data access is included — some providers treat raw data as an additional deliverable rather than part of the standard package.

Q: What is the typical turnaround time for a de novo genome assembly?

A: For a service provider handling a diploid genome under 3 Gb (e.g., rice, chicken, pig), 4–8 weeks from sample receipt to draft assembly is typical. Large genomes (wheat, barley, cattle) or polyploid genomes (sugarcane, strawberry, salmonids) typically require 8–16 weeks. Academic core facilities may operate on longer timelines (12–24 weeks) due to queue-based scheduling. Timelines that are substantially shorter — under 3 weeks for a large genome — should prompt a close look at what "assembly" means in that provider's workflow, as automated pipelines without manual curation produce rapid but lower-quality results. Please note that the turnaround time ranges mentioned above are for reference only. CD Genomics evaluates each project individually based on genome size, ploidy level, sequencing platform, and assembly strategy. For a project-specific timeline estimate, please contact us directly.

How CD Genomics Can Help

CD Genomics provides long-read sequencing services on both PacBio Revio and Oxford Nanopore PromethION platforms for agricultural genome projects, with de novo genome assembly services covering species from diploid crops to hexaploid wheat and from livestock to aquaculture. Haplotype-resolved genome assembly is available for polyploid crop genomes requiring phased chromosome-scale reconstruction. Downstream long-read data analysis covers genome annotation, comparative genomics, and variant calling tailored to agricultural species. A broader overview of the agricultural genomics platform is available on the services page. All services described here are intended for Research Use Only.

Research Use Only Statement

The information provided in this article is for research use only and is not intended for use in diagnostic or therapeutic procedures. CD Genomics provides sequencing and bioinformatics services for research purposes. Researchers should consult the appropriate regulatory guidelines for their specific applications.

References

1. Zhang T, Zhou J, Gao W, Jia Y, Wei Y, Wang G.. "Complex genome assembly based on long-read sequencing." Briefings in Bioinformatics, 2022;23(5):bbac305.. doi:10.1093/bib/bbac305
2. Zhao Z, Shi T.. "Haplotype-resolved assembly in polyploid plants: methods, challenges, and implications for evolutionary and breeding research." Genes, 2025;16(6):636.. doi:10.3390/genes16060636
3. Liu X, Zheng J, Ding J, Wu J, Zuo F, Zhang G.. "When livestock genomes meet third-generation sequencing technology: from opportunities to applications." Genes, 2024;15(2):245.. doi:10.3390/genes15020245
4. Cheng H, Asri M, Lucas J, Koren S, Li H.. "Scalable telomere-to-telomere assembly for diploid and polyploid genomes with double graph." Nature Methods, 2024;21:967–970.. doi:10.1038/s41592-024-02269-8
5. Cosma BM, Shirali Hossein Zade R, Jordan EN, van Lent P, Peng C, Pillay S, Abeel T.. "Evaluating long-read de novo assembly tools for eukaryotic genomes: insights and considerations." GigaScience, 2023;12:giad100.. doi:10.1093/gigascience/giad100