A reference genome for the eastern bettong (Bettongia gaimardi)

Sample collection and DNA/RNA extraction and sequencing

We used muscle tissue from a deceased male pouch young (B. gaimardi) individual collected from Mulligan’s Flat Woodland Sanctuary during population monitoring in 2014 and was frozen immediately after collection. The Sample was collected under Australian National University Animal Experimentation Ethics Committee ethics protocol A2011/017. DNA extraction used the QIAGEN Genomic Tips kit (Qiagen Catalogue # 10223) yielding 170ng/μl of DNA. Sequencing used the KAPA HyperPrep PCR free library kits (Roche Catalogue # KK8503) and two lanes of HiSeq Xten 150 bp PE sequencing (Illumina) at the Ramaciotti Centre for Genomics (UNSW, Sydney). Additional sequencing using the 10X Chromium Genomics library prep with >50 kb size selection was performed using 2 lanes of HiSeq Xten (150 bp PE) (Illumina). The sample was accessioned to the Australia Museum Accession number AM M.56404.

Genome assembly

Raw 10x data was assembled using Supernova v2.0.0 (RRID:SCR_016756) (Weisenfeld et al. 2017) with default settings and pseudohaplotype output. Haplotypes were tidied and filtered using Diploidocus v0.3.0 (RRID:SCR_021231) (Stuart et al. 2022), which filtered redundant and 100% unresolved scaffolds, and assigned the longest version of each scaffold to haplotype 1 [bettong.v1.0]. Scaffolds flagged as diploid (identical in both haplotypes) were then added back to haplotype 2 and each haplotype ran through Telociraptor v0.9.0 (https://github.com/slimsuite/telociraptor) to length-sort and rename scaffolds after modifying the ends to reveal telomeres, where appropriate [bettong.v1.1]. Scaffolds identified by Tiara v1.0.3 (Karlicki et al. 2022) as being from archaea, bacteria, prokarya or organelle were filtered, along with scaffolds flagged for exclusion or review by FCS-GX [bettong.v1.2] (Astashyn et al. 2023).

Completeness was estimated using Benchmarking Universal Single-Copy Orthologues (BUSCO, RRID:SCR_015008) v5.4.3 (Simao et al. 2015) using the mammalia_odb10 (n:9226) and vertebrata_odb10 datasets (n:3354).

Repetitive elements of the genome were identified, classified and masked using a Pawsey Supercomputing Centre Nimbus cloud machine (256GB RAM, 64 vCPU, 3 TB storage) by building a database using RepeatModeler v2.0.1 (RRID:SCR_015027) (Flynn et al. 2020) with default settings; repeats were then masked using RepeatMasker v4.0.9 (RRID:SCR_012954) (Smit et al. 2013-2015).

Genome annotation

A homology-based annotation was created independently for each haplotype using GeMoMa v1.9 (RRID:SCR_012954) (Keilwagen et al. 2019) using the annotation from nine Ensembl mammalian genomes (cow [Bos taurus], human [Homo sapiens], opossum [Monodelphis domestica], mouse [Mus musculus], Tammar wallaby [Macropus eugenii], platypus [Ornithorhynchus anatinus], koala [Phascolarctos cinereus], Tasmanian devil [Sarcophilus harrisii], wombat [Vombatus ursinus]) ( Table 1) and default settings. SAAGA v0.7.9 (https://github.com/slimsuite/saaga.git) was used to map annotated proteins onto a combined dataset of SwissProt (Edwards and Palopoli 2015, UniProt 2023) and Quest for Orthologues reference proteomes (Nevers et al. 2022) to add descriptions and extract the longest isoform per gene for completeness estimation using BUSCO v.5.4.3 (RRID:SCR_015008) in protein mode against the vertebrata_obd10 (n:3354) and mammalia_obd10 lineages (n:9226) (Simao et al. 2015).

The ‘genestats’ script (https://github.com/darencard/GenomeAnnotation) was used to obtain the average number of exons and introns and the average exon and intron length.

Genome assembly

Sequencing generated 185M reads of short read data and 149M reads of 10× Genomics data. Genome assembly with Supernova estimated a 3.79 Gb genome size (46.88X raw coverage) and assembled a 3.57 Gb genome in 38,249 scaffolds (scaffold N50=2.77 Mb) (Silver 2024). Following Diploidocus cleanup, there were 27,408 primary scaffolds (3.46 Gb) with 1,681 alternative scaffolds (3.01 Gb). Telociraptor made five inversions and trimmed one contig. Contamination removal filtered 786 scaffolds (2.00 Mb) from each haplotype. This gave a final genome size of 3.46 Gb with 26,623 scaffolds ( Table 2). The genome size is comparable to that of other marsupial genomes, including that of the closely related woylie (Bettongia penicillate ogilbyi) (Haouchar et al. 2016, Peel et al. 2021). BUSCO completeness of the final genome was 92.2% for mammalia_odb10 and over 96.9% for vertebrata_odb10 (96.8% for haplotype two) ( Table 2). Whilst the BUSCO scores suggest a highly complete genome, the use of long read data such as PacBio or Oxford Nanopore would assist in increasing the contiguity of the assembly. Additionally, 53.08% of the genome was identified as repeats, which is similar to other marsupials, including the closely related woylie (53.05%) (Peel et al. 2021) ( Table 3).

Genome annotation

Genome annotation with GeMoMa predicted 36,068 and 36,015 protein coding genes for haplotype one and two, respectively, which is a large over-estimation with other marsupials having around 20,000 protein coding genes (Johnson et al. 2018, Brandies et al. 2020). In the future, generating transcriptomes for a variety of eastern bettong tissues would likely improve the accuracy of the annotation. The annotation is highly complete with 94.1% of mammalian protein BUSCOs complete ( Table 4). The average protein length was 384.3 and 384.7 amino acids for haplotype one and two, respectively, with an average of 6.52 exons per gene ( Table 4). On average, predicted proteins were 87.8% the length of their best SwissProt/QFO hit, suggesting some fragmentation of the annotation, which might be inflating the numbers of annotated genes.

Ethical considerations

Samples were collected under Australian National University Animal Experimentation Ethics Committee ethics protocol A2011/017 (Approved 2011, expired Dec 2014). The sample was collected in during trapping in 2014.