Whole genome sequence analysis supports the genomic mosaicism of the endangered Hawaiian duck (Anas wyvilliana) Authors: Flor Hernández, Jonathon E. Mohl, Andrew Engilis, Philip Lavretsky Published: 2026 ABSTRACT Hybridization and gene flow are now recognized as pervasive evolutionary processes, leading to research foci attempting to understand proximate evolutionary outcomes from such events. Here, we compare the full genomes of a putative recent homoploid hybrid species, the Hawaiian duck, to its parental species, the Hawaiian Islands endemic Laysan duck and the Holarctic mallard. We find that the Hawaiian duck’s genome is a mosaic of genetic ancestry of the two parental taxa, with a greater contribution from Laysan duck, but also identify genomic regions harboring Hawaiian duck-derived alleles. Overall, our results support a hybrid origin for the Hawaiian duck and describe potential genomic regions that may have been evolutionarily beneficial for the rise and persistence of this young hybrid species on the main Hawaiian Islands. Finally, our study provides the first reference genomes for the endangered Hawaiian and Laysan ducks, valuable tools for future conservation efforts.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-49382-6. Introduction Hybridization is defined as the interbreeding of individuals from genetically distinct populations irrespective of taxonomic status1–3. Importantly, hybridization resulting in gene flow where genetic material moves between distinct organisms and thereby modifying gene pool(s)3 is increasingly being recognized as an important and pervasive evolutionary process4,5. In particular, the potential of interspecific interactions resulting in evolutionary adaptive scenarios or even in the speciation of novel taxa not only depends on the strength of reproductive barriers between the interbreeding species1,4,6,7, but also on the fitness level of the hybrids8. As a proposed scenario, if a hybrid lineage has increased fitness in a novel ecological niche over its parental lineages, then subsequent reproductive isolation between hybrids and parents is achievable, resulting in the rise of a separate species9,10. Examples of natural species that evolve via hybrid speciation are rare, but more examples of this process has been identified across plants11, insects12, flies13, mammals14, and birds15. In addition to hybrid speciation, signatures of genomic admixture can arise from the selection of one or few loci that influence a unique adaptive trait via adaptive introgression1,6,16,17. Distinguishing between these scenarios is essential when attempting to understand their prevalence and importance on the evolution of species’ genomes versus the potential for novel species formation16. Order Anseriformes exhibits an exceptionally high incidence of hybridization among birds, with an estimated 40–60% of species hybridizing2,18. In particular, the mallard complex (genus Anas), which is a group of 13–14 closely related taxa of waterfowl distributed around the world19,20, show high levels of hybridization because of their recent ancestry and/or lack of pre-zygotic barriers20. Through multi-locus genetic sampling of mitochondrial and nuclear DNA across 15 Hawaiian ducks, 21 Laysan ducks, and 25 mallards, Lavretsky et al.21 evaluated alternative hypotheses, such as stochastic sorting, mitochondrial capture, and genomic introgression. They were able to confirm that the Hawaiian duck (Anas wyvilliana; Koloa maoli) is best described as a young hybrid species, which resulted from a relatively recent hybridization event (~ 1000–5000 years before present) between Laysan duck (Anas laysanensis) and mallard (Anas platyrhynchos). The evolutionary history of the Hawaiian duck was difficult to reconcile because of incongruencies in sister relationships reconstructed with different markers or traits. Whereas morphology22 and nuclear DNA identify the Hawaiian duck as sister to Laysan ducks, mitochondrial DNA puts it closer to mallards20,21. Thus, the genome of the Hawaiian duck offers a unique opportunity to gain insights into the intricate process of genome admixture within a nascent hybrid lineage, while also testing whether such genomes arise because of hybrid speciation or adaptive introgression. The endangered Hawaiian duck is endemic to the main Hawaiian Islands, and it is phenotypically and ecologically distinct from all other mallard-like ducks. However, this species can be considered phenotypically intermediate, in terms of morphology and plumage, with respect to its putative parental species (i.e. Laysan duck and mallard)21,23–25. Previous morphological studies have suggested that Hawaiian ducks show intermediate intra-appendicular skeletal and sternal dimensions with respect to its putative parental species22. Additionally, plumage characteristics of Hawaiian ducks have also been found to be variable, with individuals displaying everything from Laysan duck to mallard-like traits23. These morphological intricacies are similar to other studies of putative hybrid species that often exhibit intermediate phenotypes to their parental species6,26. For example, the Heliconius heurippa butterfly is considered a hybrid species that exhibits an intermediate wing color pattern that promotes an assortative mating preference among hybrids12,27. Although such morphological traits are easily described, the genomic basis of these remains poorly understood. Towards this end, we compare whole-genome sequences from one individual each of the Hawaiian duck and its two putative parental taxa to establish whether behavioral and morphological intermediacy of the Hawaiian duck may be a result of genomic mosaicism. This phenomenon refers to the significant contribution of genetic material from both parental lineages, and it is considered a key signature of hybrid speciation17. We present the first reference genomes of Hawaiian and Laysan ducks, and combined these with the recent mallard reference genome28, to test whether the Hawaiian duck’s genome is indeed a mosaic of Laysan duck and mallard genetic diversity as predicted under a hybrid speciation scenario. Alternatively, the Hawaiian duck may simply be a case of adaptive introgression in which the genome is largely Laysan duck-derived with the exception of a few mallard-derived genes, or vice versa, with a genome largely mallard-derived except for a few Laysan duck-derived genes. Additionally, we assess whether the Hawaiian duck shows evidence of putatively novel derived genetic variation since divergence, while still retaining genomic ancestry from both parental species. Results Genome sequencing and assembly results We constructed scaffold-level reference genomes for a Laysan duck and a Hawaiian duck, employing a multi-level approach that combined two sequencing and assembly technologies (shotgun and Chicago; Supplementary Table S2). Generating approximately 510 million paired reads for both genomes, we achieved 36-fold and 71-fold coverage genomes for the Laysan and Hawaiian duck assemblies, respectively. Both assembled genomes were ~ 1.04 Gb in length and had a 41.01% GC content. Notably, both genomes exhibited scaffold N50 of 0.065 Mb, and the longest scaffold length reaching 1060 Mb (Supplementary Table S2). Further scaffolding using library reads and the Hi-Rise pipeline refined the final Laysan duck assembly to a total of 9721 scaffolds covering 1 Gb at the scaffold level, with an N50 sequence size of 0.761 Mb and a BUSCO scores of 80.5% complete, 4.2% fragmented, 0.7% duplicated, and 14.5% missing (BUSCO v3.1.0, eukaryota odb9 dataset, Supplementary Table S2). In comparison, the final Hawaiian duck assembly consisted of 3,859 scaffolds encompassing 1 Gb at the scaffold level, with an N50 sequence size of 8.035 Mb, and a BUSCO scores of 81.5% complete, 3.9% fragmented, 0.7% duplicated, and 13.9% missing (same dataset, Supplementary Table S2). The relatively lower BUSCO scores of our assemblies likely reflect both their fragmented draft status29 and the use of the broad eukaryota odb9 dataset30, which may underestimate completeness in avian genomes. These limitations should be taken into account when interpreting downstream analyses, as gene fragmentation may artificially reduce the apparent number of shared orthologs in comparisons. Both the Laysan and Hawaiian duck assemblies were compared to the recently published 444.6-fold coverage mallard genome that was ~ 1.04 Gb with a slightly lower GC content of 40.9%, and with a BUSCO completeness score of 95.4%28. Importantly, this assembly achieved a scaffold N50 of 79.2 Mb (L50 = 4) and the largest contig length of 195 Mb, highlighting the improved continuity and completeness of this reference genome relative to our Laysan and Hawaiian assemblies. The total number of predicted proteins varied by genome, with the highest count observed in the Laysan duck (24,952), followed by the Hawaiian duck (23,927), and the mallard having the lowest number (23,584). Utilizing OrthoVenn2, we identified an overlap of 21,680 protein clusters across the three ducks. Notably, the most significant overlap in protein clusters was observed between the Laysan duck and the Hawaiian duck (868), followed by the overlap between the Laysan duck and the mallard (599), and the lowest overlap was observed between the mallard and the Hawaiian duck (442) (Supplementary Fig. S1). The number of singletons (i.e., proteins that did not form a cluster and were not included in Supplementary Fig. S1) also varied across species, being 1,446, 655, and 578 for the Laysan duck, Hawaiian duck, and mallard, respectively. Finally, we note a general lack of unique protein clusters in the Hawaiian duck and mallard genomes, with the Laysan duck possessing only a single unique protein cluster (Supplementary Fig. S1). OrthoVenn comparative clustering analyses highlight shared versus lineage- specific repertories. High overlap implies conserved functional content, whereas few unique clusters suggest limited lineage-specific innovation, supporting the idea of strong genomic similarity among taxa31,32. Nevertheless, these values should be interpreted with caution, as genome fragmentation (particularly in the Laysan duck) may inflate protein counts by splitting single genes across scaffolds, partly explaining the discrepancy with BUSCO scores33. Finally, all downstream analyses were based on mapping the short-read data from the three focal species and the Tufted duck (Aythya fuligula) to the mallard chromosome-level reference genome, permitting us to partition datasets into autosomes, Z-sex chromosome, and the mitogenome. Across genomes, we attained high mapping similarity (~ 99%) with mean coverage ranging from 67x (mallard autosomes) to 91x (Laysan duck Z-sex chromosome) (Supplementary Table S3). In total, we recovered 879,691,675 (of ~ 1.04 Gbp) high-quality base-pairs, including 63,805,135 single nucleotide polymorphisms (SNPs). For mitogenomes, alignment success reached 100%, with average coverage between ~ 68,000 × and 100,000 × per individuals. Genome-wide patterns and genome painting To provide a genome-wide comparison among the Hawaiian duck and its parental species (mallard and Laysan duck), we focused on individual-level genomic patterns derived from whole-genome comparisons, including runs of homozygosity (ROH), heterozygosity at sites fixed between parental taxa, and local ancestry inference (genome painting). Runs of homozygosity (ROH) analysis revealed that the Laysan duck had 11 segments longer than 0.5 Mb, with average ROH length of 0.59 Mb, while the Hawaiian duck and the mallard had no ROH segments exceeding 0.5 Mb, consequently, the proportion of the genome in ROH > 1 Mb (FROH) was 0 for all individuals (Table 1). We then assessed ancestry using an ancestry painting approach that considers only sites at which the two parental species are homozygous for alternative alleles. Within this restricted set of informative sites, the Hawaiian duck showed higher heterozygosity on the Z chromosome (p_het = 0.43) than on autosomes (p_het = 0.35) (Supplementary Fig. S2, and Supplementary Table S4). In fact, painting the Laysan duck and the mallard ancestry revealed a genomic mosaicism for the Hawaiian duck (Fig. 1A). Ancestry painting further revealed that the Hawaiian duck shared a greater number of homozygous sites with the mallard across autosomes, and with the Laysan duck across the Z-sex chromosome (Supplementary Table S4). Because ancestry painting is based on counts of sites that are fixed between the sampled parental genomes, differences in parental homozygosity directly influence the number of informative sites available for such comparisons (e.g.,34,35). The extensive homozygosity observed in the Laysan genome, likely shaped by demographic history, reduces the number of sites at which the Hawaiian duck can share fixed states with Laysan on autosomes, despite overall ancestry being biased toward Laysan. We also note that these values represent counts of informative sites rather than proportional ancestry or measures of nucleotide diversity. Finally, we found that Hawaiian duck harbors the mallard mitogenome (Fig. 1A), supporting earlier hypotheses of a mitochondrial capture event21, though this pattern could also reflect retention of ancestral mitochondrial lineages in a hybrid genome.Table 1Run of homozygosity (ROH) summary.ROHSpeciesAutosomesZ chromosomeROH (# > 0.5 Mb)Laysan11–Hawaiian0–Mallard0–Mean ROH length (Mb)Laysan0.591–Hawaiian0–Mallard0–FROH (> 1 Mb)Laysan0–Hawaiian0–Mallard0–% genome in 0.5-1 Mb ROHLaysan0.63%–Hawaiian0–Mallard0–Fig. 1Parental contribution to the hybrid lineage. (A) Ancestry painting of Hawaiian duck’s genome based fixed sites between Laysan duck and mallard, resulting in 624,665 base-pairs (bp) across autosomes (left), 44,007 bp across the Z-sex chromosome, and 15 bp for the mitochondrial DNA (mtDNA; right). (B) Per chromosome D-statistics within the Hawaiian duck’s genome. (C) The size distribution of putative ancestry blocks from the Laysan duck (red), mallard (blue), and from unresolved ancestry (gray). Table 1Run of homozygosity (ROH) summary.ROHSpeciesAutosomesZ chromosomeROH (# > 0.5 Mb)Laysan11–Hawaiian0–Mallard0–Mean ROH length (Mb)Laysan0.591–Hawaiian0–Mallard0–FROH (> 1 Mb)Laysan0–Hawaiian0–Mallard0–% genome in 0.5-1 Mb ROHLaysan0.63%–Hawaiian0–Mallard0– Run of homozygosity (ROH) summary. Parental contribution to the hybrid lineage. (A) Ancestry painting of Hawaiian duck’s genome based fixed sites between Laysan duck and mallard, resulting in 624,665 base-pairs (bp) across autosomes (left), 44,007 bp across the Z-sex chromosome, and 15 bp for the mitochondrial DNA (mtDNA; right). (B) Per chromosome D-statistics within the Hawaiian duck’s genome. (C) The size distribution of putative ancestry blocks from the Laysan duck (red), mallard (blue), and from unresolved ancestry (gray). Estimating gene flow and ancestry blocks To characterize hybrid ancestry in the Hawaiian duck, we conducted four-taxon ABBA-BABA (D’ statistics) tests and fd statistics. These analyses are appropriate for single genomes and allow us to describe contributions from parental species to the Hawaiian duck genome50,51. Our four-taxon ABBA-BABA (D’ statistics) test revealed substantial genome-wide introgression occurring between both parental species and the Hawaiian duck at a whole-genome level (Table 2), as well as when analyzing chromosomes individually (Fig. 1B; Supplementary Table S5). Similarly, fd statistics further supported varying admixture proportions between the Hawaiian duck and its parental species. Notably, there is a higher degree of gene flow detected between the Hawaiian and the Laysan ducks at a whole-genome scale, but especially on the Z-sex chromosome (Table 3). Finally, the size distribution of putative ancestry blocks (Fig. 1C) shows that most blocks are small (~ 3 kb) with very few exceeding 50 kb. The prevalence of small blocks is consistent with recombination breaking down larger ancestry tracts over time36. This pattern is observed for both Laysan- and mallard-derived blocks, although there are generally more Laysan blocks, likely reflecting historical backcrossing with the Laysan duck, which were historically more abundant on the islands.Table 2Results from Patterson’s D test for hybridization between lineages with block Jack-knifed SE estimates and Z scores.P1P2P3Patterson’s DJack-knifed SEZ scoreAutosomesLaysanHawaiianMallard0.120.004030.42MallardHawaiianLaysan0.520.008066.44Z-sex chromosomeLaysanHawaiianMallard0.170.0198.78MallardHawaiianLaysan0.690.02923.51Table 3Mean fd values for 100-kb non-overlapping windows across the genomes.P1P2P3fd (%)fd (%)AutosomesZ chromosomeMallardLaysanHawaiian29.86 ± 27.6138.14 ± 27.82LaysanMallardHawaiian18.81 ± 9.263.32 ± 6.82 Table 2Results from Patterson’s D test for hybridization between lineages with block Jack-knifed SE estimates and Z scores.P1P2P3Patterson’s DJack-knifed SEZ scoreAutosomesLaysanHawaiianMallard0.120.004030.42MallardHawaiianLaysan0.520.008066.44Z-sex chromosomeLaysanHawaiianMallard0.170.0198.78MallardHawaiianLaysan0.690.02923.51 Results from Patterson’s D test for hybridization between lineages with block Jack-knifed SE estimates and Z scores. Table 3Mean fd values for 100-kb non-overlapping windows across the genomes.P1P2P3fd (%)fd (%)AutosomesZ chromosomeMallardLaysanHawaiian29.86 ± 27.6138.14 ± 27.82LaysanMallardHawaiian18.81 ± 9.263.32 ± 6.82 Mean fd values for 100-kb non-overlapping windows across the genomes. Phylogenetic relationships Utilizing RAxML, we constructed maximum likelihood phylogenies across non-overlapping 100 kb genomic windows, encompassing 267,102 autosomal SNPs and 21,045 Z-sex chromosome linked SNPs. Classifying whether the Hawaiian duck clustered monophyletically with the Laysan duck, the mallard or formed a distinct clade recovered a sister relationship between the Hawaiian and the Laysan duck across the majority of trees (69.8%). However, in a minority of cases (17.8%), the Hawaiian duck clustered with the mallard, and even forming a unique clade in 12.4% of trees (Fig. 2). A similar pattern emerged when examining Maximum Likelihood trees constructed across Z-sex chromosome windows where the Hawaiian and Laysan duck were sister taxa in 86.2% of trees, with 7.5% and 6.4% of trees placing Hawaiian duck as sister to mallard or forming a unique clade, respectively (Supplementary Fig. S3). Fig. 2Phylogenetic inference across autosomes to assess the origin of the Hawaiian duck. (A) The main three tested topologies for the position of the Hawaiian duck, and (B) their relative weightings and (C) average node placement distance across autosomes. Phylogenetic inference across autosomes to assess the origin of the Hawaiian duck. (A) The main three tested topologies for the position of the Hawaiian duck, and (B) their relative weightings and (C) average node placement distance across autosomes. Discussion Our study finds that the genome of the Hawaiian duck is indeed a mosaicism of Laysan duck and mallard genetic ancestry, and is consistent with the hypothesis that they represent a young avian homoploid hybrid species21. Ancestry painting revealed that the Hawaiian duck shares a greater number of homozygous sites with the mallard across autosomes and with the Laysan duck on the Z-sex chromosome (Supplementary Table S4). While these counts could superficially suggest a mallard-biased genome, complementary approaches suggest otherwise. Phylogenetic relationships across the genome consistently placed the Laysan and the Hawaiian duck individuals as sister taxa in majority of analyzed trees regardless of chromosome-type (Fig. 2). Moreover, ABBA-BABA (Fig. 1B; Table 2 and fd (Table 3) indicate modest but consistent overrepresentation of Laysan duck ancestry. Together, these ancestry-specific tests provide stronger support for a Laysan bias, whereas the raw counts of shared homozygous sited are influenced by the high homozygosity of the Laysan duck’s genome (Table 1), likely shaped by its severe bottlenecking to a population size of 11 individuals by 191137. Refined ancestry analyses further uncovered a fine-scale interchange of relatively small blocks (i.e., ~ 3 Kbp; Fig. 1C) of Laysan duck and mallard ancestry across the Hawaiian duck’s genome (Fig. 1A). This pattern is consistent with an ancient hybridization event followed by backcrossing with Laysan duck and subsequent interbreeding among hybrids34,36. Despite an overall admixed genome, 6–12% of phylogenetic trees placed the Hawaiian duck genome as a unique clade (Fig. 2). This suggests that divergent evolutionary processes may have played a role in shaping the Hawaiian duck’s genome since its hybrid beginnings. In addition to divergent selection, genetic drift could also contribute to this pattern, and sampling a larger pool of individuals might reveal that more alleles are actually shared among species. Although we conclude that the Hawaiian duck represents a hybrid lineage that is gradually differentiating from its parental taxa17, we acknowledge the limitations of using single-genome samples for population-level inferences. Thus, determining genomic regions responsible for the evolution and/or adaptation of Hawaiian ducks in Hawaii will require multi-individual genomic comparisons across the three species to ensure that results are not biased by individual genomes. Often, the evolution of a hybrid lineage occurs either via equal genomic contribution of parental species or through the adaptive introgression of a few genes from one parental species to the genome of the other1,6,16,17. For the Hawaiian duck, its genomic mosaicism suggests that both parental taxa substantially contributed to this individual’s genome and thus does not represent a case of adaptive introgression. Nevertheless, the slight bias in ancestry between Hawaiian and Laysan ducks may be a result of backcrossing at initial stages of hybridization38 as Laysan ducks historically outnumbered mallards, with the latter being a rare winter migrant to the Hawaiian Island that can disperse through ‘migratory drop-off’—individual birds that fail to complete their usual migratory route and remain in non-breeding areas39. Such migratory drop-offs remain an ongoing process with mallards and other migratory species establishing and reestablishing breeding populations in the Hawaiian Islands21,39. Alternatively, the observed bias could be a result of Laysan ducks harboring allelic diversity adapted to Island life, and thus, those similar selective pressures also exhibited on this Hawaiian duck hybrid genome, resulting in the retention of the same genetic diversity. Unlike the nuclear genome, our results are consistent with previous studies showing that Hawaiian ducks retain the mallard mitogenome20,21. Whether this pattern represents mitochondrial capture or lineage sorting remains unresolved. Given that mallards were the non-native parental taxa, retention of its mitogenome due to local adaptation seems unlikely, suggesting stochastic processes such as genetic drift could explain this pattern. Although our analyses consider Laysan ducks and mallards as the parental taxa as no other mallard-like ducks showed any contributions in previous work20,21, we cannot fully rule out contributions from other, now-extinct or unsampled duck populations. Using ancient DNA techniques to sequence historical and subfossil remains of putative Hawaiian ducks found across biodiversity collections can help further test between these scenarios. Moreover, because our dataset is based on a single individual per species, it does not capture intraspecific genetic diversity and limits the resolution of parental contributions. Finally we note that while the use of the mallard genome as reference could introduce mapping bias toward mallard-derived regions, the consistently high alignment rates (~99%) and uniform coverage across species (67x- 91x; Supplementary Table S3) suggest this effect was minimal. Thus, the recovery of extensive Laysan duck ancestry indicates that our inferences were not disproportionately driven by mallard-biased mapping. Future studies incorporating standard recombination rate estimates could test whether the observed ancestry block sizes correspond to the timing of hybridization events suggested by Lavretsky et al. (2015), providing further temporal resolution of the Hawaiian duck’s hybrid origin. Nonetheless, this study provides the first reference genomes for the endangered Laysan and Hawaiian ducks that can further ancient DNA research, as well as prove important tools in the future of their conservation. Methods Sampling, DNA extraction, whole-genome resequencing, and variant calling Tissue samples were collected from one Laysan duck (May 2016, Hawaiian Islands), a Hawaiian duck (November 2016, Hawaiian Islands), and a wild mallard (September 2016, New Mexico) (Supplementary Table S1). Wing or breast tissue was sent to Cantata Bio, LLC (Scotts Valley, CA) for DNA extraction, library preparation, and sequencing. All samples used in this study were previously collected and stored in institutional collections (Supplementary Table S1). No live animals were handled specifically for this research. All protocols associated with the original sampling were approved by the appropriate institutional committees. All methods were reported in accordance with ARRIVE guidelines (https://arriveguidelines.org) and relevant regulations. The Hawaiian and Laysan duck specimens were provided by the U.S. Fish and Wildlife Service and the National Wildlife Health Lab Honolulu, who donated them to the U.C. Davis Museum of Wildlife and Fish Biology (MWFB), which is a registered repository for ESA species from the Hawaiian Islands (MOU Letter of Authority signed 08 February 2005). These specimens and tissues are archived in the MWFB. Also, this research was conducted under IACUC approval 967486-1 (University of Texas at El Paso). Genome sequencing, assembly, and annotation All steps were done at Cantata Bio laboratories. DNA was first fragmented using the Bioruptor Pico for Illumina library creation. Two short-insert libraries were produced, one with an approximate insert size of 400 bp and the other with 500 bp, following Cantata Bio’s protocol and the Illumina TruSeq DNA PCR-free method. These libraries underwent sequencing on an Illumina HiSeq machine using paired-end (PE) 150-bp chemistry. Following, raw Illumina reads were first trimmed to remove sequencing adapters and poor-quality bases using Trimmonatic40. The short-insert sequences underwent analysis at various k-mer sizes (19, 31, 49, 75, 109), with a negative binomial model applied to optimize the assembly’s coverage and to balance repetitive and heterozygous sequences. De novo assembly was performed using Meraculous v241, setting the k-mer size at 55 and requiring a minimum k-mer frequency of 15, while employing the diploid nonredundant haplotig mode. Following de novo assembly with Meraculus, three Chicago libraries were prepared (Cantata Bio) following methods outlined in Putnam et al.42. Briefly, around 500 ng of HMW DNA underwent in vitro chromatin reconstitution and was then fixed using formaldehyde. The chromatin was treated with DpnII, with biotinylated nucleotides filling in the 5’ overhangs, and the remaining blunt ends were ligated. Post-ligation, crosslinks were undone to separate proteins from the DNA. The purified DNA underwent a process to eliminate non-ligated biotin, followed by fragmentation to an average size of approximately 350 bp. Sequencing libraries were created from this fragmented DNA using NEBNext Ultra enzymes and adapters compatible with Illumina. Biotin-labeled fragments were extracted using streptavidin beads prior to PCR amplification. Libraries were each sequenced on a single Illumina HiSeq X Lane using 150 bp PE chemistry. Subsequently, to scaffold and improve assembly, shotgun and Chicago library data was combined with the initial assembly from Meraculous as the input for the HiRise computational pipeline42. First, the shotgun and Chicago library sequences were aligned to the draft contig assembly using Burrows Wheeler Aligner v07.15 (bwa,43). Second, the Chicago read pairs were mapped within draft scaffolds to produce a likelihood model for genomic distance between read pairs, and this model was used to identify and break putative misjoins, score prospective joins, and make joins above the threshold. Finally, after aligning and scaffolding the draft assembly using the Chicago data, the short-insert sequences were used to close remaining gaps between contigs where possible. For the mallard duck genome, Omni-C libraries were meticulously prepared to enhance the genome assembly process (see28). For Laysan and Hawaiian duck genomes, we assessed contiguity using Quast44, and used BUSCO v3.1.045 with the associated eukaryota odb9 dataset to attain BUSCO scores. Finally, genomic assemblies were passed through Omics-box (https://www.biobam.com) to identify protein coding regions using Augustus46. OrthoVenn231 was employed to cluster homologous proteins and generate Venn diagrams across the three duck genomes. Variant calling Variant calling was performed across raw FASTQ files representing the three ducks (Laysan duck, Hawaiian duck and mallard) using a custom in-house Python script (Python scripts available at https://github.com/jonmohl/PopGen;47. Trimmomatic40 was used to trim or discard poor quality sequences. Quality sequence reads were then aligned to a high-quality de novo reference genome of a wild North American mallard28 (Available at: https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_030704485.1/) using the Burrows Wheeler Aligner v07.15 (bwa;43). Samples were then sorted and indexed in Samtools v1.648 and combined and genotyped using bcftools v1.6 (as part of the SAMtools package) “mpileup” and “call” functions with the following parameters “-c –A -Q 30 -q 30,” which set a base pair and overall sequence PHRED score of ≥ 30 to ensure that only high-quality sequences are retained. Runs of Homozygosity Runs of homozygosity (ROH) were calculated for each genome using VCFtools v0.1.1749 and using all possible base-pairs. ROH were filtered to include only segments > 0.5 Mb, and the number of ROH per individual, mean ROH length, and proportion of the genome in ROH > 1 Mb (FROH) were calculated. ROH were also categorized into size bins (0.5–1 Mb, 1–5 Mb, 5–10 Mb, > 10 Mb) to summarize the genomic distribution of homozygous segments. Ancestry painting We first visualized the Hawaiian duck’s ancestry blocks through ancestry painting following35 and34; https://github.com/millanek/tutorials/tree/master/analysis_of_introgression_with_snp_data). We first identified fixed genomic differences between the Laysan duck and mallard with no missing data using the script get_fixed_site_gts.rb. Following, base-pairs were thinned to be a minimum of 1 kb apart. We plotted the ancestry-informative sites in the Hawaiian duck using the script plot_fixed_site_gts.rb. If Hawaiian ducks represent a hybrid species, then we would expect them to have a mosaicism of Laysan duck and mallard ancestry as previously found with a few loci21. Note that all samples were males (genotype = ZZ), and thus the Z-sex chromosome was treated as diploid like autosomes. Although the mitogenome is haploid, this was also treated as diploid, and following38. Patterns of gene flow and ancestry blocks We used a four-taxon ABBA-BABA (D’ statistics) calculations to test for the presence of hybridization50,51. The Tufted duck (Aythya fulvigula) genome (retrieved from NCBI, https://www.ncbi.nlm.nih.gov/genome/?term=Aythya+fuligula; Kraus et al. Unpublished) was aligned to our three genomes and served as the outgroup in analyses. Whole genome ABBA-BABA estimates and jack-knifed standard errors were calculated for two topologies that included: {(Laysan, Hawaiian), mallard [Tufted]} and {(mallard, Hawaiian), Laysan [Tufted]}, using R code detailed in52; https://github.com/simonhmartin/tutorials/tree/master/ABBA_BABA_whole_genome). In addition to calculating the D’ statistics, we quantified the fd estimator using the ABBABABAwindow.py script52, and with the Tufted duck serving as an outgroup again. Both analyses were based on non-overlapping 100 kb windows, with a minimum requirement of 10 SNPs per window. Finally, we assessed the sizes of ancestry blocks inherited from distinct parental species within the Hawaiian duck following36; also see34. To do so, we estimated the frequency of ABBA and BABA sites within non-overlapping 3 kb windows. Windows in which the ratio of ABBA/(ABBA + BABA) exceeded 0.7 were identified as potential Laysan duck ancestry blocks, whereas those with a ratio below 0.3 were considered potential mallard ancestry windows within the Hawaiian duck genome. The length of each ancestry block was determined by the number of consecutive windows exhibiting the same ancestry pattern. Phylogenetic analysis We took a phylogenetic approach to understand the relationships between the Hawaiian duck and its parental species, with the Tufted duck serving as the outgroup. First, the full VCF file was thinned to include only SNPs which were 100 bp apart using VCFtools v0.1.1749; -thin 100). The SNP dataset was further filtered using bcftools (part of SAMtools v1.6; bcftools view -i 'COUNT(GT = “RR”) > 0 & COUNT(GT = “AA”) > 0') to leave only SNPs that were present at least once in our dataset as homozygous for the reference allele, and homozygous for the alternative allele, as required by RAxML v8.02653.The reduced VCF file was then partitioned into two dataset containing autosomal or Z-sex chromosome linked SNPs. For each of these datasets, VCF files were then converted to a fasta file using vcf2phylip.py54 to generate alignments. To investigate possible variations in coalescent histories between different genomic segments, we divided the fasta files into non-overlapping windows of 100 bp. Within each window, we estimated maximum likelihood trees using RAxML v8.02653, employing the GTR-GAMMA model. For each possible subtree insertion, RAxML evaluates the average node placement distance of the alternative topology measuring how well a subtree fits into larger phylogenetic context. Window-based subtrees were analyzed for whether the Hawaiian duck was sister to Laysan duck, mallard, or formed a unique clade. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1