An adult captive-bred male P. corroboree (Zoo Record ID: B50597), obtained from the conservation breeding program at Melbourne Zoo (Victoria, Australia), was used to generate the reference genome. Two additional captive animals: 1 male (B40300; the sire of the focal animal, wild-collected from unknown location in 2007) and 1 female (B40182; wild-collected from Upper Jagumba, NSW in 2007) were also sequenced using short reads. All frogs were humanely euthanized at the University of Melbourne (Victoria, Australia) using a buffered solution of 0.2% tricaine mesylate (MS-222) and decapitation (University of Melbourne Animal Ethics permit #10267). Using sterile techniques, multiple tissues including heart, liver, kidney, muscle, gonads, brain, and whole body were collected from each animal and flash frozen in liquid nitrogen. The tissues were immediately transferred to -80°C until they were sent for genome sequencing. Tissues were shipped on dry ice to the Vertebrate Genomics Laboratory (Rockefeller University, New York, USA) for nucleotide extraction and sequencing (Australian wildlife trade export permit #PWS2020-AU-001530).

For PacBio HiFi sequencing, high molecular weight DNA (HMW DNA) was extracted from a flash frozen kidney using the MagAttract HMW DNA Kit (Qiagen 67563). The tissue was stored at -80°C and kept on dry ice until homogenized with the Qiagen TissueRuptor II (Cat. No. 9002755). The DNA was quantified with the Qubit 3 fluorometer (Invitrogen Qubit dsDNA Broad Range Assay cat no. Q32850) and fragment size was assessed with the Agilent Fragment Analyzer.

After isolation, DNA was sheared using the Megaruptor 3 (Diagenode, Denville, NJ, USA) to attain a 15–20 Kb PacBio library insert size. The PacBio HiFi library was prepared with the SMRTbell Express Template Prep Kit 2.0 (PN 100-938-900) following the manufacturer’s procedure (PN 101-853-100 Version 03) and then size-selected with Pippin HT (Sage Science, Beverly, MA, USA). The PacBio library was sequenced on a Sequel IIe instrument with 8M SMRT cells and Sequencing Plate 2.0 (PN 101-820-200).

For Hi-C sequencing, libraries were prepared from liver using the Arima-HiC 2.0 kit (Arima Genomics, Carlsbad, CA, USA) following the manufacturer’s protocol. The library was then sequenced with the Illumina NovaSeq 6000 platform with 2×150 bp read length at Psomagen, Inc. (Rockville, MD, USA).

For Bionano optical mapping, HMW DNA was extracted from the kidney with the Circulomics Nanobind Tissue Big DNA Kit and fragment size was evaluated with a pulsed field gel electrophoresis (Pippin Pulse, SAGE Science, Beverly, MA). The DNA was labelled using direct labelling enzyme (DLE1) and Bionano Prep Direct Label and Stain (DLS) protocol (document number 30206) and then labels sequenced on a Bionano Saphyr instrument.

Total RNA was extracted from five tissues from the focal male (B50597: testes, brain, muscle, liver, and whole body) and four tissues from the female (B40182: ovary, muscle, liver, and whole body) using a QIAGEN RNAeasy Protect kit (cat. no. 74124). RNA quantity was determined using a Qubit 3 fluorometer (Invitrogen Qubit RNA High Sensitivity (HS) Kit (cat. no. Q32852) and RNA integrity (RIN) score determined using an Agilent Fragment Analyzer. RNA paired-end sequencing was performed on an Illumina NovaSeq 6000 machine.

Short read Illumina WGS sequencing of the presumed parents was performed using HMW DNA extracted from kidney tissue from the presumed parents (B40300, B40182) of the focal frog. The library was sequenced with the Illumina NovaSeq 6000 platform with 2×150 bp read length at Psomagen, Inc. (Rockville, MD, USA).

Assembly

The assembly was performed using the VGP Pipeline 2.0 with Hi-C phasing ³ . First, HiFi reads were screened for PacBio adapters using cutadapt 4.0 to remove any matching reads before assembly. Contig assembly was then performed using hifiasm version 0.16.1 in Hi-C phased mode to obtain two phased haplotypes. Contigs of each haplotype were then scaffolded individually using Bionano Solve 3.7.0 followed by YaHS 1.2a.

Assembly curation

Haplotype 2 had higher assembly metrics and thus was selected for manual curation and to be used as the main reference. The assembly was decontaminated using the Assembly Screen for Cobionts and Contaminants (ASCC) pipeline ( https://pipelines.tol.sanger.ac.uk/ascc; Aunin et al. in prep). Flat files and Hi-C contact maps used in curation were generated via the TreeVal pipeline ¹⁷ . Manual curation was conducted using the rapid curation pipeline documented at ( https://gitlab.com/wtsi-grit/rapid-curation; Wood et al. in prep) primarily using PretextView and HiGlass ¹⁸ with additional insights provided by JBrowse2 ¹⁹ . Scaffolds were visually inspected, and assembly errors corrected as described in Howe et al. ²⁰ . Any identified contamination, missed joins, and mis-joins were amended, and duplicate sequences were tagged and removed.

Evaluation of the final assembly

The final assembly was post-processed and evaluated with Nextflow ²¹ . DSL2 pipelines “sanger-tol/readmapping” ²² , “sanger-tol/genomenote” ²³ , and “sanger-tol/blobtoolkit” ²⁴ . The pipeline sanger-tol/readmapping aligns the Hi-C reads with bwa-mem2 ²⁵ and combines the alignment files with SAMtools ²⁶ . The sanger-tol/genomenote pipeline transforms the Hi-C alignments into a contact map with BEDTools ²⁷ and the Cooler tool suite ²⁸ , which is then visualised with HiGlass ¹⁸ . It also provides statistics about the assembly with the NCBI datasets report ²⁹ , computes k-mer completeness and QV consensus quality values with FastK and MERQURY.FK, and provides a completeness assessment with BUSCO ³⁰ .

The sanger-tol/blobtoolkit pipeline is a Nextflow port of the previous Snakemake Blobtoolkit pipeline ³¹ . It aligns the PacBio reads with SAMtools and minimap2 ³² and generates coverage tracks for regions of fixed size. In parallel, it queries the GoaT database ³³ to identify all matching BUSCO lineages to run BUSCO ³⁰ . For the three domain-level BUSCO lineages, the pipeline aligns the BUSCO genes to the Uniprot Reference Proteomes database ³⁴ with DIAMOND blastp ³⁵ . The genome was split into chunks according to the density of the BUSCO genes from the closest taxonomically lineage, and each chunk was aligned to the Uniprot Reference Proteomes database with DIAMOND blastx. Genome sequences that had no hit were then chunked with seqtk and aligned to the NT database with blastn ³⁶ . All outputs were then combined with the blobtools suite into a blobdir for visualisation.

The genome assembly and evaluation pipelines were developed using the nf-core tooling ³⁷ , using MultiQC ³⁸ , and making extensive use of the Conda package manager, the Bioconda initiative ³⁹ , the Biocontainers infrastructure ⁴⁰ , and the Docker ⁴¹ and Singularity ⁴² containerisation solutions.

Only one of the presumed parents, the male frog (B40300), was confirmed to be an actual parent. As a result, we could not utilise parent WGS data for haplotype phasing.

The curated assemblies were submitted to NCBI (BioProject PRJNA928730), and haplotype 2 was annotated by the NCBI Eukaryotic Genome Annotation Pipeline ⁴³ using RNA-Seq data from the kidney and liver from the male frog (SAMN32631236) and brain, ovaries, and whole body from the female frog (SAMN39610159). This annotated RefSeq version of this assembly has the accession number GCF_028390025.1 (PRJNA1082331).

Repeats were de novo modelled with RepeatModeler (Apptainer v. 1.2.3) ⁴⁴ and then annotated using RepeatMasker (v. 4.1.2-p1) ⁴⁵ with a concatenated library of genome-specific repeats generated from RepeatModeler and the Dfam amphibian repeat library (v. Dfam.h5) ⁴⁶ .

The assembly of a male Pseudophryne corroboree ( Figure 1) resulted in a genome that was 8.87 Gb in length across 12 chromosomes. The genome was sequenced using PacBio HiFi reads, generating a total of 230 Gb from 17,266,474 reads, providing approximately 26.9-fold coverage. Primary assembly contigs were scaffolded with chromosome conformation Hi-C data, which produced 381 Gb from 1,266,207,409 reads, yielding an approximate coverage of 44.5-fold. Specimen and sequencing information are summarised in Table 1.

Manual assembly curation corrected 280 missed joins or mis-joins and 1 haplotypic duplication, reducing the assembly length by 0.79% and the scaffold number by 11%, and increasing the scaffold N50 by 24.6%. The final assembly has a total of 3,127 scaffolds, with 2,699 gaps, and a large scaffold N50 of 846.9 Mb and a contig N50 of 6.8 Mb ( Table 2), well surpassing the VGP minimum metrics of 10 and 1 Mb, respectively ² . The snail plot in Figure 2 provides a summary of the assembly statistics, while the distribution of assembly scaffolds by GC proportion and coverage is shown in Figure 3. The cumulative assembly plot in Figure 4 shows curves for subsets of scaffolds assigned to different phyla. Most (92.3%) of the assembly sequence was assigned to 12 chromosomal-level scaffolds. These chromosome-scale scaffolds were confirmed by the Hi-C data and are named in order of size ( Figure 5; Table 3).

The estimated Quality Value (QV) of the final assembly is 58.5 (no more than 1 error per ~1Mb) and BUSCO v5.5.0 completeness of 89.8% (single = 88.3%, duplicated = 1.5%), using the tetrapoda_odb10 reference set ( n = 5,310). K-mer completeness for the combined haplotypes was 97.66% complete (aPseCor3.hap1 = 91.91% and aPseCor3.hap2 = 91.96%).

A considerable portion (76%) of the P. corroboree genome consisted of repeats, with the majority of these (27%) classified as Long Terminal Repeats (LTRs; Table 2).