13 14 Draft de novo reference genome assemblies were obtained from four North American killifish 15 species (Fundulus xenicus, Fundulus catenatus, Fundulus nottii, and Fundulus olivaceus) using 16 sequence reads from Illumina and Oxford Nanopore Technologies' PromethION platforms. For 17 each species, the PromethION platform was used to generate 30-45x sequence coverage, and the 18 Illumina platform was used to generate 50-160x sequence coverage. Contig N50 values ranged 19 from 0.4 Mb to 2.7 Mb, and BUSCO scores were consistently above 90% complete using the 20 Eukaryota database. Draft assemblies and raw sequencing data are available for public use. We 21 encourage use and re-use of these data for assembly benchmarking and external analyses. 22 23 Background 26 27 Sequencing and assembling large eukaryotic genomes is challenging [1-3]. Accuracy of 28 downstream analyses, such as variant calling and measuring gene expression, depends heavily on 29 a high-quality reference genome [4]. Fortunately, the cost of generating whole genome sequence 30 data is dropping, making it easier for individual labs rather than large consortiums to generate 31 assemblies for organisms without reference genomes [3,5,6]. Single-molecule long read nucleic 32 acid sequencing technology from Oxford Nanopore Technologies (ONT), which has been 33 commercially available since 2014 [7], has been shown to improve the contiguity of reference 34 assemblies [8] and reveal "dark regions" that were previously camouflaging genes [9]. The 35 lengths of the sequencing reads generated using this technology are limited only by the size of 36 the fragments in the extracted DNA sample [10]. The promise of more complete reference 37 assemblies is especially important for the accuracy of comparative evolutionary genomics 38 studies, as assembly fragments lead to errors in downstream synteny analyses [11], as well as SNP calling and identification of transcript features (splice junctions and exons) for 40 quantification. 41 42 Despite high error rates ~5% [12] relative to Illumina short reads ~0.25% [13] and the relatively 43 recent availability of ONT data, there have been a flurry of studies using this sequencing 44 technology. Small genomes from bacteria and viruses appear to be ideal for sequencing on the 45 ONT MinION platform [12]. The portable nature of the technology makes it appealing as a 46 resource for teaching [14,15], working in remote locations [16-18] and for investigating viral 47 outbreak public health emergencies [19-21]. However, despite the demonstrated ability to 48 achieve yields >6.5 Gb per flow cell [22], the MinION platform can be prohibitively expensive 49 for sequencing larger eukaryotic genomes. For example, 39 flow cells yielded 91.2 Gb of 50 sequence data (~30x coverage) of the human genome [23]. Sequencing of the wild tomato 51 species, Solanum pennellii across thirty-one flow cells yielded 110.96 Gb (∼100x coverage) with 52 some flow cells yielding >5Gb [24]. By contrast, following the 2018 beta release of the ONT 53 Prome...