henhao
Sign in
Molecular BiologybioRxiv
Preprint

On the path to reference genomes for all biodiversity: lessons learned and laboratory protocols created in the Sanger Tree of Life core laboratory over the first 2000 species

Caroline HowardiD, Amy DentoniD, Benjamin G. JacksoniD, Adam BatesiD, Jessie JayiD, Halyna YatsenkoiD, Priyanka Sethu RamaniD, Abitha ThomasiD, Graeme OatleyiD, Raquel Juliana Vionette do AmaraliD, Zeynep Ene GöktaniD, Juan Pablo Narváez-GómeziD, Isabelle Clayton Lucey, Elizabeth SinclairiD, Michael Andrew QuailiD, Mark BlaxteriD, Kerstin HoweiD, Mara LawniczakiD

PDFDOIData
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

Since its inception in 2019, the Tree of Life programme at the Wellcome Sanger Institute has released high-quality, chromosomally-resolved reference genome assemblies for over 2000 species. Tree of Life has at its core multiple teams, each of which are responsible for key components of the genome engine. One of these teams is the Tree of Life core laboratory, which is responsible for processing tissues across a wide range of species into high quality, high molecular weight DNA and intact RNA, and preparing tissues for Hi-C. Here, we detail the different workflows we have developed to successfully process a wide variety of species, covering plants, fungi, chordates, protists, arthropods, meiofauna and other metazoa. We summarise our success rates and describe how to best apply and combine the suite of current protocols, which are all publicly available at protocols.io.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
665
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 17 Oct 2026 · from the full text of version 1 (11 Apr 2025) · CC BY 4.0

Across 4883 species, standardised high-molecular-weight DNA extraction passed 96% of chordate and 91% of plant species, while fungi failed at 34.2%; ultra-low-input PacBio libraries rescue many low-yield samples.

Reference genomes for every described eukaryote, the Earth BioGenome Project goal, hinge on getting enough high-molecular-weight (HMW) DNA and intact RNA out of extremely varied taxa. The paper pairs a modular, openly shared protocol suite with taxon-level success rates from five years of production work, showing which routes work on the first pass and which groups still fail.

Key results

  1. ShownStandard extraction workflows give long-read-grade DNA for most chordates and plants, whereas fungi, protists, flatworms and amphipods fail much more often.

    Extraction pass rate 96% for chordates (91.2% pass, 4.0% pass for ultra-low-input (ULI) libraries, 1.1% after pooling) and 91% for plants; 85% across 2373 arthropod species; 83% for protists; 79% for other metazoa and macroalgae (Mollusca 87%, Platyhelminthes 29%). Fungi had the highest fail rate, 34.2%, and 71% of 24 amphipod species failed. · High Molecular Weight DNA Extraction; Arthropods; Protists; Other Metazoa and Macroalgae

  2. ShownShearing mostly succeeds, and the amplified ULI route rescues samples whose DNA is too scarce or too short for standard low-input (LI) HiFi libraries, though only for genomes under 1 Gb.

    Post-shearing pass rates ranged from 89% (chordates) to 67% (arthropods). Protists passed 20% via LI vs 54% via ULI; plants 80% vs 93% where both were tried; fungi, mostly g-TUBE-sheared for ULI, passed 79% of unique species. ULI can work from 20 ng of sheared DNA. · High Molecular Weight DNA Extraction; Fragmented DNA clean up; Plants; Fungi (including Lichens)

  3. ShownAverage HiFi yields were similar or higher for ULI than LI libraries and less variable within taxa, with a marked gain for fungi on Sequel IIe; output from a single library can still vary widely between cells.

    Mean yield per single-library run: Revio 48-69 Gb (LI) vs 65-74 Gb (ULI); Sequel 17-25 Gb (LI) vs 19-24 Gb (ULI). Fungi on Sequel IIe averaged 17 Gb with LI vs 23 Gb with ULI. One newt library (Lissotriton vulgaris, 24 Gb genome) gave 18-32 Gb per cell across seven Sequel IIe cells. · Long Read Sequencing

  4. SuggestedSwitching protocols rescued several recalcitrant groups: organic extraction for some plants, Nanobind for a whelk, bead beating for stony corals and a squeeze-based cell release for sponges.

    Plant Organic Extraction (POE) raised the Saxifragales pass rate from 56% with MagAttract to 100%. Nanobind gave a high-molecular-weight peak and tenfold yield for Colus gracilis, though tissue prep differed. Squeezing increased HMW DNA yield in Eunapius fragilis; FastPrep-96 bead beating improved coral disruption and extraction success (no counts given). · Plants; Molluscs; Cnidaria; Porifera

  5. ShownRNA extraction for gene annotation succeeded for nearly every taxonomic group, and the lab moved to an automatable bead-based protocol that needs less tissue.

    Near 100% success per taxon group (total RNA above the 100 ng library input), except protists at 92%; manual TRIzol was replaced by bead-based MagMAX mirVana, with input cut from 25 to 15 mg. · Developing a standardised pipeline for the…; RNA extraction

Methods, data and reproducibility

Snap-frozen samples are triaged by taxon, tissue and mass, homogenised (Powermasher, cryoPREP or FastPrep-96 bead beating), extracted with automated MagAttract by default plus taxon-specific fallbacks, sheared for PacBio LI or ULI HiFi libraries, and split for Hi-C and RNA-seq. Quality-control results and pass/fail calls from a frozen five-year dataset are summarised by taxonomic group.

  • Frozen Tree of Life lab dataset: DNA extraction, DNA fragmentation and RNA extraction results with a data dictionary (supplementary files) · eukaryotes from 41 phyla · n = 4883 species (DNA extraction file: nearly 5000 species)
  • Arthropod extractions · arthropods · n = 2374 species (1575 genera, 453 families, 52 orders)
  • Plant extractions · plants (927 vascular, 71 non-vascular species) · n = 998 species (63 orders, 166 families, 564 genera)
Methods & tools
Powermasher, cryoPREP and FastPrep-96 homogenisation, Qiagen MagAttract and Plant MagAttract HMW DNA extraction on KingFisher Apex, Plant Organic Extraction (POE), Nanobind and modified Omega Bio-Tek E.Z.N.A. protocols, Megaruptor 3 and Covaris g-TUBE shearing with SPRI clean-up, Qubit, Lunatic spectrophotometer and FemtoPulse (Genomic DNA 165 kb Kit), PacBio HiFi LI and ULI libraries on Sequel IIe and Revio, Arima Hi-C v2 kit with Illumina NovaSeq X, TRIzol and MagMAX mirVana RNA extraction; NEB Ultra II RNA Library Prep, GoaT (Genomes on a Tree) genome-size database, R 4.4.1 and Tableau 2024.2.1, protocols.io (Sanger Tree of Life Workspace)
Data & accessions
https://links.tol.sanger.ac.uk/datasets/tol-lab-data
Reagents & environment
Statistics in R 4.4.1 and figures in Tableau 2024.2.1; every protocol carries a version number on protocols.io, and retired versions stay online.
Reproducibility gaps
Protocols are said to be on protocols.io (Sanger Tree of Life Workspace), but no protocol links or DOIs appear in this text; the supplementary data-freeze files are mentioned without links, and the portal address is printed without http(s).
Limitations (6)
  • Amplified ULI libraries suffer coverage drop-outs from amplification bias, so they are restricted to genomes under 1 Gb; small-bodied species with large genomes (jumping spiders, isopods) remain hard. · stated by the authors
  • In cultured protists, co-grown bacteria can account for up to 99% of reads despite good yields, so cleaner cultures are needed from collectors. · stated by the authors
  • Yield from one library varies unpredictably between cells, complicating scale-up; the best disruption method for protists is still unknown, and ciliate chromosomes of 5-20 kb would be lost to current size selection. · stated by the authors
  • The Nanobind vs MagAttract comparison for Colus gracilis is confounded by different tissue preparation (cryoPREP vs scalpel dicing). · stated by the authors
  • Some figures disagree between sections: protist shearing pass rates are 20% LI and 54% ULI in one place but 57% LI and 83% ULI in another, and the arthropod count is 2373 species in one section but 2374 in another. · noted in this summary
  • Outcomes are observational records from a production pipeline, not controlled comparisons; several fixes rest on unquantified evidence: initial results for the sorbitol wash, one sponge species for the squeeze method and a small test set for FastPrep-96. · noted in this summary

Who should read:Read this if you need long-read-quality DNA, Hi-C or RNA from an unusual eukaryote and want to know which homogenisation, extraction and library routes have worked for related taxa.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGupta '25Driguez '21Shirasawa '23Montoliu-Ne '20Larivière '24Minich '23
52% semantic matchBioinformatics
Anshu Gupta, Siavash Mirarab, Yatish Turakhia·Proceedings of the National Academy of Sciences·May 2, 2025· 16 citations