Assembly

The data sources and sequential steps used to generate the genome assembly are outlined in Figure 1. These include HiFi sequencing, the use of existing optical mapping and chromosomal conformation data generated in the production of mRatBN7.2, reference-assisted assembly of subchromosomal scaffolds into single chromosomes (Alonge et al. 2022), assembly of existing CLR data into contigs for use in assembly gap filling, polishing of the assembly with Illumina short reads, and rapid curation of the assembly using Hi-C heat maps to visualize and correct misassemblies (Howe et al. 2021a).

Figure 1.

Genome assembly pipeline and data inputs. Illumina short-read sequencing was generated by the Medical College of Wisconsin (MCW) and used for assembly polishing. PacBio CLR reads generated by the Wellcome Sanger Institute and originally used for creating the preceding reference assembly mRatBN7.2 were assembled with Canu, and the resulting contigs were used for assembly gap filling. Scaffolding was performed using Bionano Genomics optical reads and Arima Genomics Hi-C data generated at Sanger for the mRatBN7.2 assembly. PacBio HiFi reads were generated on a Sequel IIe instrument at the University of Louisville. These were assembled with HiCanu, and the resulting contigs were integrated with the Bionano optical map to create a hybrid assembly using Bionano Solve software at Texas A&M University. The Hi-C data were used to further scaffold the Bionano hybrid assembly using SALSA2 software. Reference-assisted chromosome assembly used RagTag software, and the existing mRatBN7.2 assembly served as a template to order and assemble scaffolds, into which the PacBio CLR contigs were incorporated for gap filling using the RagTag patch module. After polishing, the assembly was curated using the Sanger Institute's rapid curation pipeline, which uses Hi-C contact mapping to identify and resolve misassemblies.

Open in new tabDownload PowerPointLink to figure

Our assembly of the Brown Norway genome has resulted in chromosome-level assembly with a total extent (assembly assigned to chromosomes, unlocalized and unplaced scaffolds) of 2.8496 Gb. Of this, 98.7% was assigned to chromosomes, with a scaffold N50 of 137 Mb and a scaffold L50 of 8. The comparisons reported in the present paper are between the GRCr8 assembly and the mRatBN7.2 primary assembly haplotype. We examined the HiFi reads to identify a mitochondrial genome (ChrM) within the reads using MitoHiFi software (Uliano-Silva et al. 2023). This yielded a mitochondrial chromosome that was identical to that of mRatBN7.2 except that the new assembly contained a two base insertion immediately after position ChrM:5175 of the mRatBN7.2 assembly.

The assembly was analyzed by the NCBI Foreign Contamination Screen (FCS) tool and Genome Cross-species aligner (GX) to determine whether the assembly was contaminated with extrataxonic sequences (Astashyn et al. 2024). None were found. We applied Merqury analysis to characterize aspects of the assembly using k-mer analysis (Rhie et al. 2020). Figure 2 indicates the k-mer multiplicity of the assembly as determined by Merqury analysis. The initial k-mer multiplicity plot revealed two k-mer multiplicity peaks at 20 and 40. These peaks reflect k-mer counts associated with the sex chromosomes (the subject animal was male) and the autosomes, respectively. However, the presence of residual heterozygosity in this inbred strain might lead to autosomal k-mers in the 20-peak. To assess residual heterozygosity, we extracted all reads from the BN HiFi data set that mapped to the autosomes of the GRCr8 assembly. We then generated a k-mer multiplicity profile from these reads. Plotting the coverage depth against the k-mer frequency we saw a maximum at the expected 40× coverage. The presence of heterozygosity would result in k-mers with a coverage maximum of 20×. Neither a k-mer peak nor shoulder was observed in this region. We also examined k-mer spectra derived from BN HiFi reads mapped to Chr X and Chr Y, respectively, and demonstrate that the shoulder seen in the combined histogram is caused entirely by them (Supplemental Fig. S1). Merqury also identifies k-mers that are found only in the assembly and not in the reads, indicating assembly errors; none were observed. Finally, Merqury provides an estimate of assembly completeness by comparing the total number of k-mers in reads versus the total number in the assembly. Completeness was estimated at 100%. Assembly completeness was also assessed using conserved universal single copy orthologs (BUSCO) of the Glires (rodents and lagomorphs) clade (Waterhouse et al. 2018). The recently released software compleasm was used for this assessment (Huang and Li 2023). This employs protein-to-genome alignment for the assessment of completeness. Both mRatBN7.2 and the current assembly were estimated to be 99.7% complete as represented by the inclusion of highly conserved single-copy orthologs in the assembly.

Figure 2.

We extracted all 21-mers from the HiFi reads to evaluate the k-mer distribution in the assembly. k-mers were aligned to the autosomes, extracted, and plotted to demonstrate multiplicity of k-mers. As anticipated, autosomally mapped k-mers show a single peak with a multiplicity similar to the sequencing read depth, indicative of the absence of heterozygosity. Similar mapping and plotting was done for the sex chromosomes, which both show a multiplicity indicative of their haploid state in the male subject genome. Supplemental Figure S2 shows each plot (autosomes, Chr Y, and Chr X) independently as well as the combined plot of all k-mers, allowing closer examination.

Open in new tabDownload PowerPointLink to figure

Identification of new components of the assembly

The introduction of HiFi sequencing to rat genome assembly indicated the possibility that regions of the genome with known low complexity and a high degree of repetitive sequences that were previously inaccessible might now be incorporated into the assembly. This was the case for a HiFi-based HiCanu assembly of the haploid human cell line CHM13, derived from a hydatidiform mole (Nurk et al. 2020). This assembly resolved nine of 23 expected centromeric regions. To assess the incorporation of centromeric regions in the rat assembly, we used StainedGlass software (Vollger et al. 2022) to reveal repetitive sequence in chromosomal regions expected to harbor centromeres, based on the published rat karyotype (Hamta et al. 2006). Unlike the mouse, which possesses only telomeric centromeres, distribution of rat centromeres are telocentric, acrocentric, and (sub)metacentric (see Table 1). Table 1 also indicates whether a distinct centromeric structure was detected by the StainedGlass analysis. In general, metacentric and submetacentric chromosomes had readily detectable centromeric repeat sequences that ranged in extent from 0.5 to 2 Mb (Fig. 3; Table 1). Centromere identification was more challenging in the acrocentric and telocentric chromosomes. Additional analysis of the centromeric and other noncentromeric repetitive regions added to this assembly is provided in Supplemental Figure S3. We also analyzed telomeric regions to determine whether the assembly had captured telomeric hexamer repeat sequences (Table 2). Telomeric hexamer repeats were indeed detected and found to be extensive in the metacentric chromosomes as well as in the distal terminals of acrocentric and telocentric chromosomes. Hexamer repeats were less often detected in the telomeres of telocentric chromosomes. Altogether this indicates that our assembly approach has been successful in incorporating a large amount of highly repetitive genome sequences that were not similarly captured in prior assemblies.

Figure 3.

We used the StainedGlass software tool to assess self-to-self similarity in regions of the chromosome in which the rat karyotype indicates the likely location of centromeric satellite sequences. The darkest red colors indicate near 100% identity, and purple colors indicate identity up to 90%. For each analysis, the distribution of self-similarity across the region is indicated in the graph relating number of alignments observed with the specified similarity. For illustration, we show regions of three metacentric chromosomes in which clear self-similarity is present. For more details on other chromosomes, see Table 1.

Open in new tabDownload PowerPointLink to figure

Table 3 lists the assembled chromosomes and compares their extent with those of the prior reference assembly mRatBN7.2. Notably all chromosomes have increased in size; however, the increases are not uniform in scale. Some chromosomes have increased more notably. Most obvious among these is Chromosome Y, for which only limited representation (∼18 Mb) was present in mRatBN7.2, which is discordant with Chromosome Y size expectations from the karyotype (Hamta et al. 2006). Four other chromosomes have increases in size of >10%: Chromosomes 3, 11, 12, and 19. In the case of Chromosomes 3, 11, and 12, the additional sequence is largely contributed near the proximal telomere. The acrosomal arms of these chromosomes are the location of the nucleolus-organizing regions of the rat genome, which are composed of multiple copies of ribosomal DNA genes (Tantravahi et al. 1979; Sasaki et al. 1986). These regions continue to pose a challenge to even the latest assembly methods (Rautiainen et al. 2023). We used SyRI software and the related plotsr visualization package (Goel et al. 2019; Goel and Schneeberger 2022) to analyze the syntenic relationships and structural differences between the new assembly and the mRatBN7.2. Figure 4 shows the synteny relationships among the five chromosomes in which the largest relative increases in size were observed. Supplemental Figure S2 provides similar images for the remaining chromosomes and also indicates inter-chromosomal exchanges between the assemblies. The sequence added to chromosomes in this new assembly generally occurs in contiguous blocks, often corresponding to the expected centromere locations (Hamta et al. 2006). In addition, small inversions are noted, as are several translocations and duplications. SyRI also provides a summary and the extent of various types of structural variation between mRatBN7.2 and the current assembly (Table 4). The “inversions,” “translocations,” and “duplications” noted in this table refer to segment-wise differences between two reference assemblies; they are not differences between individual animals.

Figure 4.

Synteny analysis was performed with SyRI software to illustrate the shared regions of chromosomes with the largest percentage increase in size in the present assembly compared with mRatBN7.2. Chr 3, Chr 11, and Chr 12 are acrosomal chromosomes in which rDNA arrays are located in the acrosomal arms. The current assembly has greatly extended the chromosome in the region of these arms. Chr 19 shows a large insertion spanning >15 Mb that was not included in mRatBN7.2. Chr Y shows good synteny with mRatBN7.2 across the Y-specific coding regions of the chromosome. However, the more distal regions were not assembled in mRatBN7.2, undoubtedly reflecting the repetitive heterochromatic region of this chromosome.

Open in new tabDownload PowerPointLink to figure

We have examined the incorporation of new protein-coding genes into the assembly by using NCBI annotation data. These fall into two categories: 780 unmapped genes that could not be placed on the mRatBN7.2 assembly via whole-genome alignments but are localized to chromosomes in GRCr8, and 373 novel genes whose sequence is present in both assemblies but were not previously annotated on mRatBN7.2 by NCBI . A listing of these genes is provided in Supplemental Table 1. A majority of both categories of genes, including 76% of the unmapped and 60% of the novel genes, share homology with known genes. Many of the unmapped genes appear to be related to testis and sperm function and localize to repetitive regions of the genome in which expansion of these genes has taken place.

The pseudoautosomal regions of the sex chromosomes

When meiotic recombination between the X and Y Chromosomes occurs, it is limited to the pseudoautosomal region (PAR). In humans and other primates, the X and Y Chromosomes are (sub)metacentric. However, in murids (including rats and mice), the sex chromosomes are telocentric. Recent studies suggest that the PAR in murids has undergone progressive degeneration with loss of PAR genes either to autosomal locations or completely from the genome. We investigated in our assembly the chromosomal location of genes present in the human PAR. In the mouse (Mus musculus, Mmu), only two human PAR gene orthologs were identified in the mouse PAR (Maxeiner et al. 2020). Studies of a prior R. norvegicus genome assembly did not identify any PAR genes in the proximal ends of the sex chromosomes (Maxeiner et al. 2020). We examined the location of orthologs to genes located in the human PAR in our assembly. As is shown in Table 5, we confirmed the absence of prospective PAR genes in proximal sex chromosomes in our assembly, finding that such genes either were absent from the assembly or are located on autosomes. This suggests that recombination capacity mediated by PAR of the X and Y Chromosomes in humans may be absent in the rat and that the telomeric region of the rat sex chromosomes may have a function limited to chromosomal synapsis and chiasma formation.

Gene annotation

The mRatBN7.2 assembly was adopted by the GRC in November 2020 as the rat reference assembly, with the Rat Genome Database (RGD) as a GRC member for its ongoing curation and updates (Vedi et al. 2023). It has undergone thorough annotation using methodologies and data sets established at NCBI and Ensembl. GRCr8 has recently undergone annotation by the RefSeq group at NCBI. The annotation used about 10 billion short reads and about 35 million long reads and incorporates curated RefSeq transcripts (NM_ prefixes) for ∼80% of coding genes. We compared the recent June 2023 mRatBN7.2 NCBI annotation report (GCF_015227675.2-RS_2023_06) with the annotation report of GRCr8 (GCF_036323735.1-RS_2024_02). Notable advances are observed across most assessments, including the annotation of 1100 protein coding genes absent from the prior annotation. Table 6 provides a comparison and includes equivalent data for the current mouse reference assembly GRCm39 (NCBI Mus musculus annotation release 109). Although the mouse assembly annotation arises from a different data set of mouse RefSeq sequences, which includes a larger accumulation of cDNA and other sequencing data for mouse, it provides some indication of how the current rat reference annotation has advanced to a similar level as the current mouse reference. The prior rat reference assembly has more than 200 annotated genes that require additional curation to resolve discrepancies. Of these, 17 genes in the present assembly had the same problem as in mRatBN7.2, implying an origin in the RefSeq gene rather than the assembly. We used isoform sequencing (Iso-Seq) alignments to investigate 33 remaining genes with problematic concordance with the annotation gene set in GRCr8. Of these, 24 appeared to be problems arising in the annotation gene set, as the assembly was substantiated with concordant alignments. There were five genes that could not be investigated because no Iso-Seq alignments were present. Of the four remaining genes, one was not located in the assembly, two appear to reflect gaps in the assembly, and the final one was an MHC gene identified in a rat strain known to have a different MHC haplotype from the reference.

We used Liftoff software (Shumate and Salzberg 2021) to determine to what extent annotations (NCBI, UCSC, and Ensembl) of mRatBN7.2 could be transferred to the new assembly. This would provide an indication of the extent to which the new assembly had failed to incorporate gene coding regions into the assembly. Supplemental Table 2 indicates that only a handful of genes could not be transferred using this tool.

To gain further insight into the expressed genome, we used PacBio Iso-Seq data from multiple unique tissues of individuals from the same BN strain used for the assembly. These data are helpful to validate the NCBI annotation and provide further insight into potential novel isoforms. We have also evaluated the utility of these data to inform about gene expression from newly incorporated segments in this GRCr8 assembly. To examine this region further, we again used StainedGlass software to provide visualization of self-similarity across the region from Chr 19: 9 Mb–35 Mb. As can be seen in Figure 5, this revealed an extensive region of related repetitive sequences on Chromosome 19 from 14 Mb to 30 Mb.

Figure 5.

Alignment of Iso-Seq data from several BN tissues obtained from the MCW suggested that the newly incorporated region on Chr 19 shows aligned transcript data across its extent. This is in contrast to newly incorporated centromeric regions in which no aligned transcripts were observed. To investigate further, we examined whether the included region might be composed of expanded genomic sequences with high similarity, such as might occur as the result of repeated segmental duplication. Self-to-self similarity visualization with StainedGlass reveals a highly duplicated structure that lacked the simple expansion observed in centromeric regions. This newly incorporated region of Chr 19 contains 276 previously unmapped RefSeq genes and 12 additional proposed genes from Gnomon annotation analysis.

Open in new tabDownload PowerPointLink to figure

Base-level assembly accuracy

Merqury analysis of GRCr8 provided a QV estimate of the assembly of 59.5 (Rhie et al. 2020). In our evaluation of mRatBN7.2, we found 129,000 variants shared by more than 156 inbred rats, from a data collection containing 163 inbred rats, which includes 12 BN/NHsdMcwi samples (de Jong et al. 2024).The most parsimonious explanation for the large number of variants shared by all rats, including BN/NHsdMcwi, the reference strain, is that these are not authentic variants unique to the Brown Norway rat genome, but rather result from PacBio CLR sequencing in the creation of the mRatBN7.2 assembly which is prone to sequencing errors. Because HiFi sequencing is much more accurate than CLR, we repeated this analysis, comparing variants shared by 42 rat strains, including several BN substrains, and discovered that only 550 variants were shared in common by most strains but divergent in our assembly. This provides a useful indication of the reduced inclusion of sequencing errors in the present assembly. Although much of the present assembly approaches the recently reported telomere-to-telomere assembly contiguity and completeness (Nurk et al. 2022; Rautiainen et al. 2023), additional new sequencing data, such as can be obtained using pore-based sequencing methods, will be required to generate an assembly in which remaining complex structural elements are resolved.

Whole-genome sequencing

The tissue sample used for the long-read sequencing was obtained from the liver of a male rat of the inbred BN strain aged 18 weeks. This subject animal was raised in a closed colony at the Medical College of Wisconsin with strain name BN/NHsdMcwi and was registered at the RGD under the identifier RGD_61498. High-molecular-weight (HMW) genomic DNA (gDNA) was isolated using a MagAttract HMW DNA Kit (Qiagen 67563) following the manufacturer's protocol. HiFi and CLR WGS SMRTbell libraries were generated as per manufacturer's instructions (Pacific Biosciences). SMRT sequencing was performed on the Sequel IIe instrument (Pacific Biosciences) using 2.0 chemistry and four SMRT cells. Following data collection, circular consensus sequencing (CCS) analyses were performed to generate highly accurate intramolecular consensus reads. Those CCS reads that demonstrated an intrinsic read quality of >Q20 (99%) were classified as “HiFi reads” and used for downstream assembly, resulting in 40-fold HiFi read coverage. We also used publicly available long-read CLR sequence data (European Nucleotide Archive accession ERR5310326 and ERR5310327) that was produced for the preceding rat reference genome assembly, mRatBN7.2 (Howe et al. 2021b), to provide gap filling information for the assembly. This sequence came from a male animal from the same strain and colony as used for the HiFi reads. Short-read sequences (150 bp end) with ∼50× coverage depth (TruSeq, Illumina HiSeq X) were also obtained from a BN male of the same strain and colony sequenced at the Medical College of Wisconsin (Sequence Read Archive IDs SRA: SRS17979924; SRA: SRS17979921; and SRA: SRS17979833). Short-read whole-genome sequencing data from 42 rat strains were obtained from NIH SRA (Supplemental Table 3) and were aligned to the assembly to validate base-level accuracy of the assembly.

Genome assembly

Assembly using HiFi reads employed HiCanu software (version 2.1.1), with an expected genome size set to 2.7 Gb (Nurk et al. 2020). The command line option -pacbio-hifi was used to specify the HiFi sequence FASTA files for assembly.

Assembly scaffolding

Two methods were employed to provide long-range information suitable for scaffolding the HiCanu-generated contigs. Optical genome mapping data were obtained from publicly available data (European Nucleotide Archive accession ERZ1741012) generated from the prior rat reference genome assembly. A hybrid assembly was created from the HiCanu contigs and a de novo optical map. Creation of the de novo map and its integration with the HiCanu contigs used Bionano Saphyr Solve software (v3.3) to generate hybrid scaffolds. Publicly available Arima Hi-C chromosome conformation capture data produced for the prior rat reference genome assembly (European Nucleotide Archive accession ERR5309023 and ERR5309024) were used to scaffold the hybrid scaffold assembly using SALSA2 software (Ghurye et al. 2019).

Chromosome builds, gap filling, and polishing

We evaluated and corrected scaffolds and integrated subchromosomal scaffolds into full-length chromosome builds using RagTag v2.1.0 software (Alonge et al. 2022). The existing NCBI rat reference assembly mRatBN7.2 (GCA_015227675.2) (Howe et al. 2021b) was used to template chromosome-level scaffolding. To increase joining gaps in the assembly, we used the “patch” module of RagTag in combination with the CLR data. The CLR reads were assembled with Canu software, and the resulting contigs were provided to RagTag for alignment-guided gap filling using minimap2 (Li 2018) to align the CLR reads to the unpatched assembly. Polishing of the assembly may reduce sequence inaccuracy arising from microsatellite and homopolymer repeat errors that can occur in HiFi-based assemblies and may improve accuracy in regions that were gap-filled with CLR-based contigs. Polishing can be done because the Brown Norway rat cannot homozygously differ from itself. In principle, any homozygous differences identified between its sequence data and the reference derived from them is an error and can be corrected. This was done with Pilon version 1.24 (Walker et al. 2014), where BN short-read WGS was mapped to the assembly. The memory required by Pilon was reduced by breaking the assembly up into individual scaffolds for analysis using the SAMtools faidx command (Li and Durbin 2009). Single-scaffold mapped data sets were extracted using SAMtools view, and each scaffold was processed through Pilon individually and aggregated back into a single assembly after polishing. It should be noted that the polishing step was run prior to the inclusion of the mitochondrial genome in the assembly. As such, reads that would have mapped to ChrM mapped to the nuclear-embedded mitochondrial (NUMT) DNA sequences instead. In some cases, this would have resulted in Pilon editing the nuclear genome to match alleles found in the mitochondrial genome. These artifacts along with the other homozygous differences that remain between the assembly and sequence data from the same strain after polishing are included as Supplemental Data tables in a tab-delineated text format (Supplemental Table 4).

Assembly annotation

Annotation of the assembly was performed by NCBI using standardized methods of the NCBI pipeline for reference genome annotation (Thibaud-Nissen et al. 2013). We also examined the ability of annotations from mRatBN7.2 to be transferred to the GRCr8 assembly using Liftoff software (Shumate and Salzberg 2021). New protein-coding genes in GRCr8 were identified using NCBI annotation files based on GeneIDs not found in the mRatBN7.2 annotation and were subdivided into “current-unmapped” and “current-novel” categories based on NCBI annotation comparison files, in which “unmapped” refers to genes with no comparable location on mRatBN7.2 found through assembly alignments and “novel” refers to genes with a comparable location but no gene was previously annotated.

Alignment of full-length BN/NHsdMcwi gene transcripts to the assembly

Samples were collected from 19 tissues (lung, spleen, liver, kidney, heart, brain, skin, stomach, ileum, colon, prostate, ovary, epididymal fat, urinary bladder, thymus, skeletal muscle, mixed vascular tissue, nucleated blood cells, and testis) from 18 week old BN/NHsdMcwi animals (one female from the same colony provided ovary; all other tissues came from the same male animal from which the HiFi sequencing was obtained). Tissues were placed in RNAlater until further use for long-read Iso-Seq preparation. Nucleated blood cells were purified by density centrifugation using SepMate reagents (Stem Cell Technologies) and transferred to RNAlater. RNA was extracted from ∼25 mg of tissue using the RNeasy Mini Kit (Qiagen), as recommended by the manufacturer. The quality and quantity of the RNA were assessed by a Bioanalyzer 2000 (Agilent) and Qubit fluorometry (Thermo Fisher Scientific), respectively. Barcoded Iso-Seq SMRTbell libraries were generated as recommended by the manufacturer (Pacific Biosciences) and sequenced using 2.1 chemistry and 30 h movies on the Sequel IIe system. Prior to sequencing, RNA samples were multiplexed in batches of four tissues, and each fourplex pool was sequenced using a single SMRTcell. Following data collection, HiFi reads were generated on the instrument and used as input into the Iso-Seq v3 workflow within the SMRT Link v10.0 online SMRT sequencing bioinformatics suite. Briefly, the Iso-Seq v3 pipeline trims barcodes, primers, and poly(A) tails; reorients transcripts; and filters out PCR concatemers, producing de novo predicted transcripts as “full-length nonconcatemer” (FLNC) reads that were used for downstream analyses. The BAM files generated by SMRT Link data processing of Iso-Seq data have been deposited at NCBI and are linked to the NCBI genome assembly project (NCBI PRJNA1027884).

The FLNC reads were converted to FASTA files and aligned to the genome assembly using minimap2. The resulting FLNC sam file was converted to BAM format using SAMtools, sorted, and indexed for viewing in Integrative Genomics Viewer (IGV) software (Robinson et al. 2011). The aligned Iso-Seq data have been used to investigate regions of the genome that were absent from the prior reference assembly to determine whether gene expression arises within these regions and to examine which tissues generate such reads.

Assembly evaluation

Completeness of the assembly was assessed using BUSCO, using compleasm software (Huang and Li 2023). We compared the new assembly with the prior assembly to identify regions of synteny or structural and base-level differences between them using SyRI software (Goel et al. 2019; Jiao and Schneeberger 2020; Goel and Schneeberger 2022). Base-level accuracy of the assembly was also investigated by comparison of the sharing of gene variants between other inbred rat strains and the GRCr8 assembly. Base-level accuracy of the assembly was also investigated by identifying genomic variants shared by all four BN/NHsdMcwi samples and at least 34 out of additional 38 inbred strains of rats. Short-read data from these samples were mapped against the GRCr8 assembly using BWA (Li 2013); variants were discovered using DeepVariant (Poplin et al. 2018) and jointly analyzed using GLnexus (Yun et al. 2021). Gor a list of SRA read records of these rat strain sequences, see Supplemental Table 3. We used StainedGlass software (Vollger et al. 2022) as a tool to identify centromeric structures within the assembly. This search was guided by prior studies of the rat karyotype (Hamta et al. 2006) as well as evidence from genome-to-genome alignment suggesting locations of newly incorporated sequences in the region of potential centromeres (see Fig. 4, Chr 19). We also searched the assembly to localize and measure the presence of telomeric hexameric repeats using the seqkt program (https://github.com/lh3/seqtk).