In 2016, a team of researchers posted a draft genome for the salmon louse, Lepeophtheirus salmonis, a marine copepod that costs the global aquaculture industry hundreds of millions of dollars each year. The draft was fragmented, unannotated, and largely unusable for the kinds of population-level studies that would have made it valuable. For years, it sat in public databases, cited by almost no one. Then, a Norwegian lab decided to rebuild it.

The story of that rebuild is a lesson in how scientific data becomes useful, and how a reference genome, once assembled with care, can transform a field. It is also a reminder that a genome draft is not an endpoint but a starting point, and that the quality of that starting point determines what downstream science is possible.

A Genome Draft No One Cited

The 2016 draft was assembled from short-read sequencing data, which produced thousands of small contigs. Without a genetic map or chromosome-scale scaffolding, the assembly was a jumble of pieces that were difficult to align and even harder to interpret. The annotation, where genes are identified and named, was minimal. For a researcher interested in, say, the genetic basis of drug resistance in the louse, the draft offered little beyond a rough inventory of sequences.

Citation counts for the draft were near zero. A handful of papers mentioned it in passing, but few attempted to use it as a primary resource. The reasons were practical. Short-read assemblies are prone to gaps and misjoins, making them unreliable for variant calling. Without a clear chromosomal context, researchers could not map quantitative trait loci or track allele frequencies across populations. The draft was, in effect, a pile of puzzle pieces with no picture on the box.

One of the few labs that did try to work with it was at the Norwegian Veterinary Institute, where researchers had been studying salmon louse biology for years. They needed a reference that could support genotyping-by-sequencing and RNA-seq experiments. What they found instead was a scaffold that collapsed in places and expanded in others, with thousands of small contigs that could not be placed. The draft was not just incomplete; it was misleading.

A few early users attempted to annotate genes themselves, but the effort was enormous. Without a curated reference, every analysis required extensive filtering and manual curation. The result was that most labs simply avoided the draft altogether, opting instead to design their own reduced-representation sequencing approaches that did not require a full genome. The draft became an orphan, a dataset that was technically public but effectively invisible.

Consider the experience of a graduate student at the University of Bergen in 2018. She spent three months trying to align RNA-seq reads from louse salivary glands to the draft, only to find that a significant fraction of reads mapped to multiple locations or to nothing at all. The repetitive regions, which are common in copepod genomes, were so poorly assembled that the alignments were meaningless. She eventually gave up and switched to a transcriptome assembly, which captured only expressed genes and missed the regulatory regions. Her story was typical. The draft was not just a nuisance; it was a barrier to progress.

Why a Reference Matters for a Parasite

The salmon louse is an external parasite that feeds on the skin and mucus of farmed salmon. Infestations cause stress, open wounds, and secondary infections, and they cost the industry an estimated half a billion dollars annually through treatment costs and lost production. Control relies on a narrow set of chemical treatments, and resistance to those treatments has been spreading for decades. Understanding the genetic basis of that resistance is essential for developing new control strategies.

A high-quality reference genome is the foundation for that work. It allows researchers to identify genes involved in detoxification, to track the spread of resistance alleles, and to compare populations from different farms and regions. It also enables studies of the louse's evolutionary history, including how it has adapted to the artificial environment of fish farms.

The 2016 draft was too fragmented for these tasks. Variant calling requires a reference that is both complete and accurate. If the reference has misassembled regions, then every downstream analysis inherits those errors. For a parasite with a large genome and a high repeat content, the risk of misassembly is especially high. Short reads alone cannot resolve the long, repetitive regions that are common in many genomes.

For years, researchers in the field made do with partial information. They used transcriptomes, which capture only expressed genes, or they used reduced-representation approaches that sample a fraction of the genome. These methods can answer some questions, but they are blind to structural variation and to the regulatory regions that often drive resistance. As one researcher put it, they were working with a map that had entire counties missing.

To appreciate the stakes, consider the economics of salmon farming. Norway alone produces over a million tonnes of farmed salmon each year, and sea lice are the single largest health challenge. A single treatment for sea lice can cost a farmer tens of thousands of dollars, and the industry spends roughly a billion kroner annually on delousing measures. Resistance to emamectin benzoate, the most widely used treatment, was first reported in the late 2000s and has since spread along the entire Norwegian coast. Without a reliable reference genome, efforts to understand and manage that resistance were severely hampered.

One Lab's Rebuild: The Methodical Pivot

The Norwegian team, led by geneticist Rasmus Skern-Mauritzen, decided to rebuild the reference from scratch. Their approach was methodical and deliberate. Instead of relying on short reads, they used long-read sequencing, which produces reads of tens of thousands of bases and can span repetitive regions that short reads cannot. Long-read technology was still relatively new in 2018, but the cost had dropped enough to make the project feasible.

The team also used genetic maps to scaffold the long-read contigs into chromosome-scale groups. By crossing lice with known markers, they could order and orient the contigs along the louse's chromosomes. This step was critical, because it turned a collection of contigs into a set of chromosomes, each with a defined structure. The result was a draft that was not only more contiguous but also more accurate.

The rebuild took time. The team had to generate new sequencing data, develop custom pipelines for assembly and scaffolding, and then carefully curate the result. They also had to annotate the genome, identifying genes and their functions. This required integrating multiple lines of evidence, including RNA-seq data from different life stages and tissues. The process was iterative, with each round of annotation revealing new issues to fix.

One of the biggest challenges was dealing with the repetitive elements. The salmon louse genome is estimated to be around 700 megabases, and a substantial fraction consists of transposable elements and other repeats. Long reads can span some of these repeats, but not all. The team had to develop sophisticated algorithms to resolve the most complicated regions, and even then, some gaps remained. They chose to leave those gaps rather than risk misassembly, a decision that prioritized accuracy over completeness.

Another challenge was the genetic map. To construct it, the team had to rear lice in the laboratory, cross them, and genotype hundreds of offspring. This is no small feat for a marine parasite that requires a fish host. The process took months and required careful husbandry. But the effort paid off, as the genetic map provided the framework for ordering the contigs into chromosomes.

By 2021, the team had produced a near-complete reference genome. The assembly was anchored to chromosomes, with most genes placed in their correct positions. The annotation included thousands of gene models, many of which had been missing or incorrect in the 2016 draft. For the first time, the salmon louse had a reference that could support serious population genetics and functional genomics.

From Uncited to Indispensable

When the new draft was posted on a preprint server in late 2021, the response was immediate. Researchers who had struggled with the old draft began using the new one within weeks. Citation counts, which had been near zero for years, jumped within months. The preprint was downloaded hundreds of times, and soon became the standard reference for any study involving the salmon louse genome.

The impact was most visible in population genetics. With a reliable reference, researchers could now call variants across the genome and compare allele frequencies between populations. Studies began to appear that mapped regions associated with resistance to emamectin benzoate, a common treatment. These regions contained genes known to be involved in detoxification, such as cytochrome P450s and ATP-binding cassette transporters. The reference made it possible to identify the specific alleles that confer resistance and to track their spread.

For example, a 2022 study used the new reference to compare lice from farms with high and low treatment success. They found a small number of single-nucleotide polymorphisms that were strongly associated with resistance, and several of these fell within or near genes that are known to metabolize xenobiotics. The study also revealed that resistance alleles were not confined to a single population but had spread across the entire coast, likely through the movement of infected fish. This kind of analysis was simply impossible with the old draft.

The reference also enabled studies of the louse's evolutionary history. By comparing the genome to those of other copepods, researchers could identify genes that have undergone positive selection, including those involved in host-parasite interactions. The reference became a tool for understanding how the louse has adapted to life on farmed salmon, and how it might continue to evolve in response to new control strategies.

Beyond resistance, the reference has been used to study the louse's immune system. Researchers have identified a family of genes that encode proteins similar to those found in vertebrate immune systems, suggesting that the louse has a more sophisticated immune response than previously thought. This could have implications for vaccine development, a long-sought goal in the industry. The reference has also been used to assemble the louse's mitochondrial genome, which is useful for tracing maternal lineages and population history.

The contrast with the 2016 draft was stark. Where the old draft had been an obstacle, the new one was an enabler. Papers that had previously been impossible to write were now being published. The reference had become indispensable, and the field was moving faster than ever before.

The Lessons for Genome Publishing

The salmon louse story offers several lessons for genome publishing. The first is that drafts without curation are dead ends. A genome assembly is not a finished product; it is a working hypothesis that must be validated and refined. The 2016 draft was posted with minimal annotation and no chromosome-scale scaffolding, and it was effectively unusable. Posting such a draft may satisfy a data-sharing mandate, but it does little to advance science.

The second lesson is that funding must cover assembly validation. Building a genome is expensive, but so is validating it. The Norwegian team's rebuild required long-read sequencing, genetic mapping, and extensive manual curation. None of that work is glamorous, but it is essential. Without it, the genome is just a pile of data. Funding agencies that require genome projects to include validation and curation would save money in the long run by avoiding the need for costly rebuilds.

The third lesson is that deprecated references need clear flags. When a new, improved reference is published, the old one should be clearly marked as superseded, with a pointer to the new version. This is not always done, and researchers may unknowingly use an outdated reference, wasting time and effort. The salmon louse community was fortunate that the new reference was widely publicized, but not all communities are so vigilant.

Finally, rebuilds deserve recognition, not reinvention. The Norwegian team did not start from nothing; they built on the foundation of the 2016 draft, even if that foundation was shaky. Their work was an act of curation and improvement, not a fresh start. Yet in the current publishing culture, such rebuilds are often undervalued. They are seen as incremental, not novel. This is a mistake. Rebuilds can be as important as original assemblies, and they deserve the same recognition.

Consider the analogy of a library. The first draft is like a pile of unbound pages dumped on a table. It contains information, but it is not usable. The rebuild is like a librarian who sorts the pages, binds them into volumes, and catalogues them. Without the librarian, the library is a mess. The librarian's work is often invisible, but it is what makes the library functional. Similarly, genome curators are the unsung heroes of genomics, and their work should be celebrated.

What This Means for Future Parasite Genomics

The salmon louse case is a template for how to handle genomes of other parasites and pests. Many species that are economically or medically important have genomes that were assembled with short reads and are now known to be fragmented. The same long-read technologies that enabled the salmon louse rebuild are now widely available, and the cost continues to fall. There is no reason to accept a substandard reference when a better one is within reach.

Community standards for genome quality are also evolving. The Vertebrate Genomes Project and the Earth Biogenome Project have set ambitious goals for chromosome-level assemblies, and the tools to achieve them are improving. For parasites, which often have large genomes with high repeat content, long-read sequencing is particularly valuable. The salmon louse genome, for example, is about 700 megabases, larger than many other arthropods, and its repetitive regions are extensive. Short reads alone could never have resolved them.

Shared references accelerate all downstream work. Once a high-quality reference exists, every subsequent study benefits. The salmon louse reference has already been used to develop genetic markers for monitoring resistance, to study the louse's microbiome, and to explore its immune system. Each of these projects would have been far more difficult, or impossible, with the old draft.

The field is now considering which other parasites deserve a similar treatment. Sea lice in the Atlantic, for example, are a major problem for salmon farming, and their genomes are similarly fragmented. The same approach that worked for L. salmonis could be applied to them. The lesson is that a reference genome is not a one-time investment; it is a living resource that must be maintained and improved.

But there are also counter-arguments to consider. Some researchers argue that not every species needs a chromosome-level assembly, and that the cost of such projects can be prohibitive. For species that are not the focus of intensive research, a draft-level assembly might be sufficient. The salmon louse, however, is a species with enormous economic impact and a pressing need for genetic tools. The investment in a high-quality reference was justified by the potential return.

Another counter-argument is that the time and effort spent on rebuilding a genome could be better spent on functional studies. Why not just use RNA-seq or reduced-representation sequencing to answer the immediate questions? The answer is that those approaches are limited. They cannot capture structural variation, which is often the key to understanding resistance. They also cannot provide the context needed to interpret non-coding variants. A reference genome is a foundational resource that enables a wide range of studies, and its value grows over time.

None of this is to say that the salmon louse rebuild was easy, or that every genome project should aim for the same level of completeness. There are trade-offs between cost, time, and quality, and not every species needs a chromosome-level assembly. But for species that are the focus of active research, the salmon louse case shows what is possible when a lab decides to do the meticulous work of curation. The 2016 draft was not a failure; it was a beginning. The rebuild turned it into a foundation.