The Genomics Revolution in Nonmodel Species: Predictions vs. Reality for Salmonids

ABSTRACT

The increasing feasibility of whole‐genome sequencing has been highly anticipated, promising to transform our understanding of the biology of nonmodel species. Notably, dramatic cost reductions beginning around 2007 with the advent of high‐throughput sequencing inspired publications heralding the ‘genomics revolution’, with predictions about its future impacts. Although such predictions served as useful guideposts, value is added when statements are evaluated with the benefit of hindsight. Here, we review 10 key predictions made early in the genomics revolution, highlighting those realised while identifying challenges limiting others. We focus on predictions concerning applied aspects of genomics and examples involving salmonid species which, due to their socioeconomic and ecological significance, have been frontrunners in applications of genomics in nonmodel species. Predicted outcomes included enhanced analytical power, deeper insights into the genetic basis of phenotype and fitness variation, disease management and breeding program advancements. Although many predictions have materialised, several expectations remain unmet due to technological, analytical and knowledge barriers. Additionally, largely unforeseen advancements, including the identification and management applicability of large‐effect loci, close‐kin mark–recapture, environmental DNA and gene editing have added under‐anticipated value. Finally, emerging innovations in artificial intelligence and bioinformatics offer promising new directions. This retrospective evaluation of the impacts of the genomic revolution offers insights into the future of genomics for nonmodel species.

1. The Promise of Genomics to Change Everything

As the human genome project progressed through the 1990s and into the 21st century, discussion intensified regarding how the impending ‘genomics revolution’ would revolutionise molecular studies of nonmodel species. Initial enthusiasm surrounding genomic sequencing was catalysed by the potential to enhance analytical power, provide deeper insights into functional genetic diversity and improve quantitative studies of biodiversity. With anticipated applications in ecology and evolution (Wolfe and Li 2003; Snape et al. 2004; Medina 2005), conservation and management (Hedrick 2001; Luikart et al. 2003; Wilson et al. 2005), and agriculture and aquaculture (Meuwissen et al. 2001; Goddardt 2003; Varshney et al. 2005), accessible genomic data promised to transform scientific research.

During this period of anticipation (pre‐2007), costs of genomic sequencing largely followed Moore's law, which postulated a constant rate of increasing computational output with decreasing cost (Figure 1; Moore 1965; Wetterstrand 2024). Yet, for nonmodel species and studies requiring genomic data for many individuals or populations, large‐scale sequencing remained cost prohibitive. Instead, alternative technologies—for example, microsatellites and SNP assays—facilitated development of bioinformatic pipelines and analytical tools for processing relatively sparse marker‐based data (Selkoe and Toonen 2006; Shen et al. 2009). Following the advent of 454's massively parallel pyrosequencing protocol (Margulies et al. 2005) and Illumina's reversible terminator chemistry (Bentley et al. 2008), next‐generation sequencing (NGS) prices rapidly outpaced Moore's law beginning around 2007 (Figure 1), making whole‐genome sequencing widely accessible for researchers operating beyond the scope of model species. Subsequently, a number of perspective articles were published, heralding the arrival of the genomics revolution and outlining specific predictions for how genomics would improve research capacities across disciplines (Table 1).

FIGURE 1.

The decline of short‐read sequencing costs (lines) and the rise of methodological innovations (stars) advancing the field of population genomics and allowing novel applications in nonmodel species. Lines give ‘actual’ (red) and ‘theoretical’ (Moore's law, blue) costs of sequencing. Cost per megabase (Mb, left y‐axis) sourced from Wetterstrand (2024), cost per salmon genome (right y‐axis) assumes continuous 1x coverage across a 2.5 gigabase (Gb) genome. Stars indicate the year in which major innovations were published and became widely accessible to nonmodel researchers; some methods may have been implemented before first publication. Note, long‐read technologies are increasingly utilised and associated sequencing cost per Mb may differ from short‐reads.

TABLE 1.

Summary of predictions and outcomes.

Prediction	Predicted by	Subjective summary of realized predictions	Subjective summary of caveats for predictions	Empirical examples	Perspectives and reviews
1. Improved estimation of key demographic parameters	Luikart et al. (2003), Ouborg et al. (2010), Shafer et al. (2015)	Genetic‐based estimates of demographic parameters like N c , N e , N b , and migration rate have become more accurate and precise		Lehnert et al. (2019), Kovach et al. (2020), Fraik et al. (2021), King et al. (2023), Kurland et al. (2024), Bekkevold et al. (2024), Goetz et al. (2024), Liu et al. (2024), Mobley et al. (2024), Rosenbaum et al. (2024)
2. Increased analytical power through access to more markers	Luikart et al. (2003), Primmer (2009), Allendorf et al. (2010)	Tens of thousands to millions of SNPs are now routinely used, which has resulted in increased analytical power	There is a distinct limit to how much utility is gained in many applications, leading researchers to revert to low‐density panels	Hauser et al. (2011), Meek et al. (2016), Meek et al. (2020), Gao et al. (2023), Hargrove et al. (2024)	Flanagan and Jones (2018), Waples et al. (2020)
3. Enabled large‐scale kinship studies for insights into inbreeding and fitness traits	Luikart et al. (2003), Allendorf et al. (2010), Avise (2010), Steiner et al. (2013)	Targeted marker panels have enabled cost‐effective, long‐term pedigree reconstruction of natural populations. Wild pedigrees are offering new insights to eco‐evolutionary dynamics, impacts of supportive breeding, and inbreeding depression. Increased analytical power has increased the accuracy of genomic estimates of kinship and inbreeding	A missing but crucial advancement in salmonids is in understanding effects of inbreeding on fitness and population dynamics. Although, genomic data have vastly improved understanding of inbreeding depression in other species (e.g., Soay sheep, red deer, and killer whales)	Christie et al. (2012), Robinson et al. (2017), Christie et al. (2018), May et al. (2020), Shedd et al. (2022), Beulke et al. (2023), May et al. (2023), Dayan, Sard, et al. (2024), McPhee et al. (2024)	Flanagan and Jones (2018)
4. Expanded number of study species and populations	Luikart et al. (2003), Allendorf et al. (2010), Ouborg et al. (2010), Ekblom and Galindo (2011), Richards (2015)	Genetic data has enabled cost effective monitoring of a greater number of populations and nonmodel species	Funding still dictates research priorities; economically valuable species tend to be overrepresented. Still, genomic resources developed in commercially valuable species may be useful in resource‐limited, conservation settings	Dayan, Mazur, et al. (2024), Pappas et al. (2023)	Evans et al. (2016), Formenti et al. (2022)
5. Enhanced detection of adaptive genetic diversity	Luikart et al. (2003), Primmer (2009), Allendorf et al. (2010), Ekblom and Galindo (2011), Steiner et al. (2013), Shafer et al. (2015), Richards (2015)	The identification of adaptive loci is now essentially routine. Selective breeding programs use genomic selection to leverage adaptive genetic variation and enhance production traits. In rare cases, marker assisted selection is used to select for or against specific alleles at QTLs; gene editing to replace unwanted alleles may become more widespread	Separation of adaptive loci from false positives remains challenging. Conversely, the extent of false negatives is likely extremely high, especially for highly polygenic traits. Using adaptive loci to inform assisted migration or other management actions has been limited. There are few cases where adaptive loci have been useful for conservation. In many cases, gene‐targeted conservation is cautioned against. Genomic offsets offer a potentially promising path forward but remain unvalidated in many use cases	Apgar et al. (2017), Leitwein et al. (2017), Willoughby et al. (2018), Pearse et al. (2019), Willis et al. (2020), Thompson et al. (2020), Czorlich et al. (2022), Harder and Christie (2022), Tigano and Russello (2022), Andrews et al. (2023), Clare et al. (2023), Euclide et al. (2023), Kurland et al. (2024), Raunsgard et al. (2024), Rougemont et al. (2023), Aykanat et al. (2024), Howe et al. (2024), Hugentobler et al. (2024), Lecomte et al. (2024), Miettinen et al. (2024), Narum et al. (2024), Sparks et al. (2024), Thompson et al. (2024), Tigano et al. (2024), Willis, Coykendall, et al. (2024), Willis, Stephenson, et al. (2024)	Pearse (2016), Kardos and Shafer (2018), Rellstab et al. (2021)
6. Effective detection, monitoring, and treatment of disease	Steiner et al. (2013), Richards (2015)	Genomic information has increased precision of QTL detection, which has played a major role in marker assisted selection and genomic selection for disease resistance in aquaculture. Genomics and transcriptomics have also enabled advanced testing for disease in wild populations, enhancing monitoring efforts	Our ability to manage, prevent, and treat disease in natural systems remains extremely limited	Moen et al. (2015), Tsai et al. (2016), Vallejo et al. (2017), Boison et al. (2019), Robledo et al. (2019), Purcell et al. (2018), Mordecai et al. (2021), Vallejo et al. (2024), Palti et al. (2024)	D'Agaro et al. (2021), Robinson et al. (2023)
7. Utility and rise of population‐level transcriptomics, functional genomics, metabolomics, epigenetics	Primmer (2009), Ouborg et al. (2010), Ekblom and Galindo (2011)	Gene expression methods are being used to examine gene function and metabolic pathways. Transcriptomics are enabling novel therapeutic development for disease treatment in aquaculture. These methods have become hugely important in annotating functional genomics	While population‐level variation in gene expression exists, applications to conservation and management are limited. The utility of epigenetics is promising but remains to be seen	Tomalty et al. (2015), Baerwald et al. (2016), Christie et al. (2016), Christensen et al. (2020), Kokkonen et al. (2024), Paris et al. (2024)
8. Genomic data may create more challenges than it solves	Allendorf et al. (2010), Steiner et al. (2013), Shafer et al. (2015)	Adoption of advanced computational and informatic methods was initially slow due to limited availability of user‐friendly software, pipelines, and knowledge gaps	Bioinformatic resources are now largely keeping pace with increases in genomic data, allowing new challenges to be met. Few examples exist where genomic data impedes policy actions. Knowledge gaps between science and policy have likely not increased dramatically	Lowry et al. (2017), Bertho et al. (2022), Dimos and Phelps (2023)	Hemstrom et al. (2024)
9. Traditional marker types will remain useful	Primmer (2009), Shafer et al. (2015)	Older marker types like microsatellites are still routinely used in many management applications, although results may not be published as often	Genomic methods are generally preferred by researchers even when traditional methods are sufficient. Slow transition to ‘omics may be due to financial barriers. Work from non‐academic, management‐focused labs may be underrepresented in the literature’	Kaland et al. (2023)	White et al. (2023), Meek and Larson (2019)
10. Improved delineation of conservation units	Allendorf et al. (2010), Steiner et al. (2013)	Increased analytical power has facilitated the accurate description of finer‐scale population structure and species delimitations, improving conservation decisions	Opinions differ regarding the appropriate scale for managing natural populations (e.g., geographical and/or gene‐specific levels)	Funk et al. (2012), Kovach et al. (2022)	Waples and Lindley (2018), Waples, Ford, et al. (2022), Lehnert et al. (2023), Tengstedt et al. (2024)

The most recent decade has witnessed the generation of previously unthinkable quantities of sequencing data. Technological and methodological advancements enabling this rapid increase (Figure 1) were accompanied by significant progress in computational power and bioinformatic methods, enabling the analysis of vast amounts of genomic data and facilitating advanced biological inference. The rise of reproducible scientific practices coupled with open access bioinformatic tools has, in some ways, democratised access to robust genomic analysis pipelines, improving possibilities for researchers worldwide to contribute to and benefit from these advancements (Hemstrom et al. 2024; Hogg 2024). In practice, research funding and publishing are still dramatically inequitable, with few well‐funded laboratories, predominantly in the global north, generating a vastly disproportionate amount of genomic research (i.e., Linck and Cadena 2024). Reflecting on outcomes from the genomics revolution situates us to better evaluate research and funding priorities for the coming decades.

Here, we examine prevalent predictions made during the onset of the genomics revolution, specifically concerning applied research of nonmodel species. Although this revolution has fundamentally advanced our understanding of basic biological processes, we focus primarily on predictions concerning applications of genomics for conservation, fish and wildlife management, and agriculture. Fulfilled predictions showcase the utility of increased analytical power and improved understanding of adaptive diversity enabled by genomic data. Conversely, challenges and barriers have prevented certain predictions from being fully realised, including both technological limitations and biological realities, knowledge gaps and the rapid obsolescence of certain methods.

2. Salmonids as a Case Study

To sharpen the focus, we review the genomics revolution through the lens of Salmonidae (hereafter, salmonids), a family of fishes comprising approximately 70 to  > 200 species (Dysin et al. 2022), including salmon, trout, char, grayling and whitefish. Salmonids are the most widely studied family of nonmodel fishes, are forerunners in genomic applications, and thus can be used as a case study for identifying lessons learned across taxa (Garner et al. 2016; Waples et al. 2020). In fact, between 2010 and 2018, more than 20% of all conservation genetics studies on fish species were salmonid‐focussed, nearly five times more than the next most represented family of fishes (Bernos et al. 2020), in stark contrast to the limited research on species of lesser human interest (Mouquet et al. 2024). The extensive research focus on salmonids reflects broad, interdisciplinary interest in this taxon for fisheries management, large‐scale commercial aquaculture operations, strong cultural significance and vastly disproportionate funding for conservation of at‐risk natural populations (Evans et al. 2016). Salmonids have thus become an emerging model in ecological and evolutionary genomics (Waples et al. 2020). Salmonid researchers were early adopters of NGS and genomic approaches, and resulting studies comprise a diverse range of genomic applications. Throughout the text and in Figure 1, we primarily cite literature focussed on salmonids; however, we include examples from other nonmodel taxa in the few cases where salmonids lack relevant examples. While several salmonid species have been studied much more extensively than typical ‘nonmodel’ organisms, the vast majority of salmonids lack significant research investment, and their poor characteristics as laboratory animals (i.e., long generation time and expensive system requirements) prevents them from being researched to the same extent as Drosophila, Mus musculus , Arabidopsis and other classic model species. Interestingly, substantial investments in genomic resources in a few, well‐studied species (i.e., O. mykiss and S. salar ) have benefitted research advancements in the entire Salmonidae family (Figure 2)—a trend that can be expected in other nonmodel groups.

FIGURE 2.

Species within Salmonidae that have at least one publicly available, high‐quality (scaffold‐level or higher) genome assembly on NCBI as of March 2025. Genera with reference sequences are organised into a simplified phylogeny following Wang et al. 2022. Most represented species only have a single available genome sequence; however, several species have multiple distinct assemblies (count denoted in parentheses). Notably, Salmonidae represents at least 70 species (Dysin et al. 2022), most of which, including several high‐profile salmonids (i.e., ESA‐listed Bull trout, Salvelinus confluentus ), still lack a genome assembly and thus are not displayed above.

3. Fulfilled Predictions—With Some Caveats

The genomics revolution has delivered on many of its anticipated outcomes, broadly transforming our understanding and management of nonmodel species (Figure 3A). One of the most extensively fulfilled predictions has been improved estimation of demographic parameters such as census size, effective population size, effective number of breeders and migration rates (Table 1, Prediction 1). Genomic studies have leveraged a massive increase in the number of markers (Table 1, Prediction 2) to improve the accuracy and precision of demographic estimates (Table 1, Prediction 1), despite important limits to the utility of additional markers (Box 1). Improved estimates of demographic parameters have enhanced identification of at‐risk species (e.g., Kovach et al. 2022) and delineation of population boundaries (e.g., Funk et al. 2012). Furthermore, genome‐wide data have opened the door to advanced coalescent approaches to examine complex, long‐term demographic histories (Mather et al. 2020; Table 1, Prediction 1), as well as close‐kin mark–recapture (CKMR) methods that leverage kinship to examine recent demography (Table 1, Prediction 3; discussed further below). The ability to reliably and cost‐effectively reconstruct long‐term pedigrees of wild populations—primarily using small, targeted marker panels (i.e., May et al. 2020; Mobley et al. 2024)—has provided insights into eco‐evolutionary dynamics (Galla et al. 2022), outcomes of supportive breeding (Ivy and Lacy 2010), and delineation of fine‐scale population structure (May et al. 2023). However, the genomic basis of and effects of inbreeding on fitness and population dynamics in salmonids remain relatively unexplored compared to other taxa (Harrisson et al. 2019; Stoffel et al. 2021; Kardos et al. 2023; Hewett et al. 2024). Nevertheless, increased precision in estimating these parameters, for example by examining runs of homozygosity (Yoshida et al. 2020), supports more informed conservation strategies and management practices (Kardos et al. 2016; Kurland et al. 2024) for wild and captive‐bred salmonid populations.

FIGURE 3.

A graphical abstract depicting predictions vs. reality of the genomics revolution for salmonid research.

BOX 1. Limits to Increased Precision in Genomics‐Scale Datasets.

Modern genomics technology can now routinely generate 10³–10⁷ single‐nucleotide polymorphisms (SNPs), even for nonmodel species. Each diallelic SNP contains information about one allele, with the other being determined by subtraction. Therefore, each SNP potentially adds one degree of freedom to the estimate of a genetic parameter, with the overall number of degrees of freedom (n) used to set confidence intervals around the point estimate. If all SNPs were independent, the variance of estimates of key genetic metrics like F _ST would be inversely proportional to the number of SNPs (L): $\mathit{var}\left({\widehat{F}}_{\mathit{ST}\left(L\right)}\right)=2{\left[E\left(F_{\mathit{ST}}\right)\right]}^2/L$ (Lewontin and Krakauer 1973), where ${\widehat{F}}_{\mathit{ST}\left(L\right)}$ is the mean estimate based on information from L SNPs, and the number of degrees of freedom associated with this estimate is n = L.

In reality, physical linkage (loci occurring close together on the same chromosome) creates lack of independence (pseudoreplication) among SNPs (Waples et al. 2022), and this effect becomes more pronounced as L becomes large compared to the number of chromosomes (Chr). As a consequence, the marginal benefit (in terms of increasing precision) of adding an additional locus declines as the number of loci already in the analysis grows larger. The marginal benefit can be quantified as the rate of increase in the effective degrees of freedom (n’). For analyses that focus on linkage disequilibrium (LD) at pairs of loci, a second type of lack of independence also is important: overlapping pairs of the same loci. For L loci there are L(L‐1)/2 ≈ L ²/2 different pairs of loci, but only L/2 of these pairs (fraction ≈1/L) are completely non‐overlapping. For the LD method for estimating N _e , the nominal degrees of freedom associated with an estimate using L SNPs is n = L(L−1)/2 (Hill 1981), but n’ is reduced both by physical linkage and this overlapping‐pairs effect.

The magnitude of these effects on precision for ${\widehat{F}}_{\mathit{ST}}$ and ${\widehat{N}}_{e\left(\mathit{LD}\right)}$ is illustrated in Figure A1, which shows that n’ is strongly affected not only by L but also by true N _e and by genome size (indexed here by chromosome number, Chr). Several points are worth noting: (1) Effects of pseudoreplication are much more pronounced for analyses based on pairs of loci (bottom panels): After a few thousand SNPs are used, adding more loci has a negligible effect on precision for ${\widehat{N}}_{e\left(\mathit{LD}\right)}$. In contrast, unless genome size or true N _e is very small, precision for F _ST can continue to increase well beyond 50,000 SNPs (top panels). (2) Pseudoreplication is less of a limitation when true N _e is large (left panels). This is a propitious result for N _e estimation, as it is very challenging to obtain robust estimates for large populations. (3) Pseudoreplication is less of a limitation when genome size is large (right panels). For perspective, Drosophila have 4 chromosomes, while 25 is the mean chromosome number for vertebrates (Li et al. 2011).

FIGURE A1.

Marginal increase in effective degrees of freedom (n’) as a function of the number of diallelic loci already included in the analysis (X axes), for estimates of F _ST (top) and of N _e using the LD method (bottom). The Y axes show the expected increase in n’ from adding 10 more loci to the analysis. n’ is used to set confidence intervals for ${\widehat{F}}_{\mathrm{ST}}$ and ${\widehat{N}}_e$. In theory, n increases linearly with the number of loci for F _ST and proportional to the square of the number of loci in the LD method, but lack of independence causes the actual rate of increase in n’ to be slower. The curves shown are based on results from Waples, Waples and Ward (2022), assuming samples of S = 50 individuals. In the left panels, the number of chromosomes was fixed at Chr = 20, and true N _e was 100 (solid pink lines) or 1000 (dotted blue lines). In the right panels, N _e was fixed at 200, and the number of chromosomes was 4 (solid pink lines) or 25 (dotted blue lines). Results for the LD method assume that calculation of r ² is restricted to pairs of loci on different chromosomes. Note the order of magnitude difference in scale of the X axes in the top and bottom panels.

Another significant realised prediction has been expansion in the number of populations and species that can be effectively monitored due to substantial decreases in sequencing costs (Table 1, Prediction 4). Although genetic research on salmonids is still dominated by the most economically valued species (i.e., Atlantic salmon, S. salar ; sockeye salmon, O. nerka ; Chinook salmon, O. tshawytscha ; rainbow trout, O. mykiss , and brown trout, Salmo trutta ), substantial work is also underway investigating less commercially‐lucrative species, particularly those of conservation concern including Alpine whitefish (Coregonus sp., ‘balchen’), brook trout ( S. fontinalis ), bull trout ( S. confluentus ), cutthroat trout ( O. clarkii ), European grayling ( T. thymallus ), gila trout ( O. gilae ) and lake trout ( S. namaycush ; De‐Kayne et al. 2020; Mamoozadeh et al. 2023; Amish et al. 2022; Kovach et al. 2022; Sävilammi et al. 2019; Camak et al. 2021; Smith et al. 2022). The increasing availability of high‐quality reference genomes for diverse species has also increased the accessibility of genomic approaches for more species and populations (Formenti et al. 2022). Instead of undertaking a de novo genome assembly project, researchers can now routinely rely on previously published assemblies for their species of interest or closely related species (Figure 2).

Enhanced detection of adaptive genetic diversity has also been a prominent realised prediction (Table 1, Prediction 5), particularly associated with disease resistance (Steiner et al. 2013). Detection of putatively adaptive loci in natural populations is seemingly now routine (Table 1, Prediction 5), although the lack of consistency across studies and analysis methods remains a concern (Hemstrom et al. 2024). More recently, the detection of large‐effect loci (discussed further below) has shifted the field to a focus on multi‐omic, genome‐to‐phenome (G2P) studies. This shift has been particularly noticeable in aquaculture research, where genomic architectures of resistance to diseases such as infectious pancreatic necrosis virus (Moen et al. 2015), infectious haematopoietic necrosis virus (Miller et al. 2004; Palti et al. 2024) and sea lice (Tsai et al. 2016; Robledo et al. 2019; Vallejo et al. 2024) have been characterised and incorporated into breeding programs through genomic selection, and in some cases, frequencies of deleterious variants have been reduced through marker‐assisted selection (Table 1, Prediction 6). Notably, complete purging of deleterious alleles is difficult and slow even in selective breeding contexts (i.e., Götz et al. 2015) and is unlikely to happen in most conservation settings. California condors, for example, have been selectively bred for decades to mitigate a lethal recessive allele causing chondrodystrophy, yet the deleterious variant remains in the population (Robinson et al. 2021).

Unlike in aquaculture settings, the practical implications of analyses of adaptive diversity are less clear for natural populations. ‘Genomic offsets’, the estimated difference in genomic composition between a contemporary population and one optimally adapted to an anticipated future environment, show promise to identify vulnerable populations and harness adaptive genomic diversity to benefit assisted migration, climate risk assessment and conservation prioritisation. However, further empirical validation of genomic offset approaches is needed, and concerns exist over the frequency of false adaptive signatures (Rellstab et al. 2021, Lachmuth et al. 2023, May et al. 2025 in review). Additionally, detection and monitoring of diseases in nature has improved (Storfer et al. 2021) and important functional genomic variation has been identified, such as in the case of large‐effect loci controlling spawning migration‐timing, age at maturity, and ecotypic morphology (Table 1, Prediction 5). However, although several examples exist where large‐effect loci provide relevant knowledge for management (e.g., greb1L, six6, vgll3, Jensen et al. 2022; Waples, Ford, et al. 2022), including structural variations such as chromosomal inversions (Pearse et al. 2019; Hale et al. 2024), few examples exist where large‐effect loci have been directly incorporated into management (Waples et al. 2020) and integration of adaptive genetic variants into routine monitoring remains rare (but see Miettinen et al. 2024).

4. Challenges and Unmet Expectations

Despite important advancements, several predictions remain unfulfilled or have met challenges that warrant examination (Figure 3C). One example is the predicted utility and rise of population‐level transcriptomics, metabolomics, proteomics, and epigenetics (Table 1, Prediction 7). These fields, the combined use of which are now broadly referred to as ‘multi‐omics’ (Subramanian et al. 2020), have indeed proven instrumental in annotating gene function, providing a mechanistic understanding of biological systems (e.g., genome‐to‐phenome), and are leading to novel therapeutic developments for disease treatment in aquaculture (Suravajhala et al. 2016). However, while the use of multi‐omic approaches in research is increasing, direct applications in conservation and management are largely lacking, and their potential future role remains unclear. In addition, despite increasing recognition of the importance of noncoding variation, enhancer elements, and DNA conformation, the practical utility of these insights for conservation and management strategies remains underdeveloped.

Another largely unmet prediction was that genome‐wide data might create more problems for conservation and management than it solves (Table 1, Prediction 8). Perhaps unsurprisingly, the adoption of advanced genomic methods in management and policy has lagged behind the recent rapid availability of genomic data. This discrepancy was initially due to a scarcity of user‐friendly software, standardised pipelines, and knowledge gaps between academics and practitioners (Shafer et al. 2015; Taylor et al. 2017; Klütsch and Laikre 2021). However, the development of bioinformatic resources and improvements in science‐to‐policy communication have significantly mitigated these issues (Garner et al. 2016; Hogg et al. 2022; Hemstrom et al. 2024; Mirchandani et al. 2024). Despite this progress, integrating genomic data into established management practices remains complex (Kardos and Shafer 2018; Hickey et al. 2019; Waples et al. 2020). Although there are few documented instances where genomic data have directly hindered conservation actions, there are cases where the rapidly growing field of genomics has introduced complexity to conservation conversations, delaying decision‐making processes (Waples, Ford, et al. 2022). The gap between scientific advancements and policy implementation appears to persist. However, this gap has narrowed considerably over the past decade, and it should be noted that management‐focussed laboratories may simply not publish their reports in peer‐reviewed journals as often as academics, which does not preclude their use of the latest technologies (Garner et al. 2016). While knowledge gaps are frequently cited in grant applications and discussed as barriers, they may now reflect funding needs more than true logistical challenges. Overcoming remaining obstacles to fully harness the potential of genomic data will require continued efforts to validate bioinformatic tools and approaches to data analysis and interpretation, enhanced practitioner training and more effective communication between scientists, managers and policymakers. And yet, instances exist where improved genomic applications cannot be expected to solve particular conservation problems. For example, habitat quality, predation by invasive species, and interspecific competition can be more limiting to population recovery than genetic diversity or other intrinsic genomic factors.

In some cases, predictions were fulfilled, but not necessarily for anticipated reasons. Such is true of the continued use of traditional marker types in many management applications (e.g., microsatellites). Although it was predicted that these older marker types would not become obsolete due to their established utility (Table 1, Prediction 9), we posit that their continued use is most likely due to a desire for data continuity, user familiarity and/or economic reasons. For example, the cost and time required to ‘update’ the marker type for samples in long‐term data sets can be much greater than genotyping a smaller number of new samples with the traditional marker. The transition to ‘omics technologies that are, in many cases, cheaper and faster’ (i.e., low‐ or medium‐density marker panels) has not been universally adopted, particularly in management‐focussed laboratories where there may be a lack of resources or capacity to adopt new methods (e.g., funds and/or training). For example, the US Fish and Wildlife Service still requires microsatellite genotyping of Atlantic salmon broodstock in aquaculture settings for population assignment and parentage‐based tagging purposes (National Research Council (NRC) et al. 2004, King et al. 2005), despite availability of more accurate, versatile SNP‐based methods (Bradbury et al. 2015; Gao et al. 2023). Some laboratories that have been slow to change contend that the benefits of long‐term data consistency outweigh any costs. However, inclusion of traditional markers into new marker panels or converting traditional markers to high‐throughput sequencing (Gruenthal and Larson 2021) represents one viable solution (i.e., high‐throughput microsatellite panels; Marcy‐Quay et al. 2023). Generally, trade‐offs continue to exist in identifying the most cost‐effective approach to meet specific project objectives, given available genotyping technologies and data density requirements. However, the importance of considering these tradeoffs among various genotyping methods may diminish in the coming years, as high coverage whole genome sequencing becomes a more affordable and flexible option. While high‐density data may not be strictly necessary for certain applications (Box 1), once generated, it can easily be bioinformatically subsetted. Moreover, additional benefits of generating high‐density data—such as improved consistency, opportunities for collaboration, and future research applications—should encourage continued transitions toward high‐density approaches as they become more cost effective.

In other cases, predictions were made about promising new technologies (not included in Table 1) that could theoretically revolutionise the field, yet these technologies were rapidly replaced by newer omics methods. For example, the utility of bacterial artificial chromosomes (BAC) libraries and Expressed Sequence Tags (ESTs; predicted by Primmer 2009, Ouborg et al. 2010) was heralded for their utility in building new genome assemblies and identifying gene‐linked markers. The rapid obsolescence of these methods might be interpreted as an unmet prediction, but only because transcriptomic and genomic sequencing was so readily adopted for these applications.

Improving the delineation of conservation units (CUs) was another predicted outcome that has seen mixed results (Table 1, Prediction 10). Increased analytical power (Box 1) has enabled robust descriptions of population structure, even at the scale of mere meters for salmon populations within small streams (May et al. 2023). However, genome‐wide and traditional marker‐based approaches often align in identifying large‐scale genetic groups, such as Evolutionarily Significant Units (ESUs). This congruence has limited genomics' impact on CU designation, though its utility is notable in detecting cryptic species and characterising adaptive genetic variation (e.g., spawning timing, maturity age, migratory tendency in salmonids; Waples, Ford, et al. 2022). Enhanced marker power has nonetheless refined and adjusted ESU boundaries in specific cases involving threatened taxa (Andrews et al. 2018; McCartney‐Melstad et al. 2018). Genomic data have also clarified taxonomy and resolved cryptic phylogenies (Coimbra et al. 2021; Gu et al. 2023; Bertola et al. 2024; Morin et al. 2024), potentially aiding in future CU delineations.

Differing opinions regarding the appropriate geographic scale for managing natural populations and the relative importance of geographic versus putatively adaptive gene‐specific variation have also tempered the role of genomics in CU designation (Funk et al. 2012, 2019; Chhina et al. 2024). In many cases, gene‐targeted conservation is cautioned against, as it may overlook broader ecological and evolutionary processes (Pearse 2016; Kardos and Shafer 2018; Kardos et al. 2021). Thus, it can be postulated that genomic‐level knowledge has created new challenges in this respect.

5. Largely Unforeseen Developments

Several mostly unanticipated but influential developments have unfolded during the genomics revolution for nonmodel species (Figure 3B). One of the most surprising outcomes has been the widespread prevalence of large‐effect loci. Numerous salmonid studies have now described the genetic architecture for traits such as adult migration timing, age at maturity, disease resistance, and ecotypic diversity, revealing an unexpectedly high proportion of phenotypic variation explained by few genomic regions (Table 1, Prediction 8). These traits were previously assumed to have largely polygenic architectures, and few researchers anticipated the frequent role that large‐effect loci appear to play in the eco‐evolutionary dynamics of nonmodel species (Waples, Ford, et al. 2022). In fact, the persistence of these large effect polymorphisms in multiple species suggests associated life history traits are likely under balancing selection. Moreover, the large‐effect loci discovered include those localised to one or two genes, as well as large genomic regions encompassing hundreds of genes linked in chromosomal inversions (Wellenreuther and Bernatchez 2018; Pearse et al. 2019; Matschiner et al. 2022), and comparative genomics analyses have uncovered striking examples of parallel evolution of inversions in genomic regions with deep synteny (MacGuigan et al. 2023). As highlighted above, the characterisation of large‐effect loci is prompting redefined conservation units in wild populations and increasing our ability to purge harmful loci in aquaculture settings through marker‐assisted selection. New efforts linking large‐effect loci to reproduction and physiology offer a promising path towards a more comprehensive, multi‐omic future (Ahi et al. 2024; Verta et al. 2024).

Another largely unforeseen development that continues to garner interest is the application of close‐kin mark–recapture (CKMR) to estimate abundance, numbers of breeders and other demographic parameters in wild populations (Skaug 2001; Bravington and Grewe 2007; Rawding et al. 2014; Bravington et al. 2016). This methodology leverages genetic markers to identify related pairs of individuals and a modified mark–recapture framework to robustly estimate census population sizes. CKMR allows for accurate and precise abundance estimation for populations that have been notoriously difficult to describe using traditional survey methods. Although improved demographic estimates were predicted (Table 1, Prediction 1), the utility of CKMR was not anticipated for both populations studied using traditional mark–recapture, as well as a suite of additional species. While early CKMR applications used microsatellites to identify parent‐offspring pairs (Pearse et al. 2001), genomics has enabled more complex relationship inference (Table 1, Prediction 3), expanding the power and reliability of CKMR, particularly in long‐lived species with iteroparous lifecycles. While further paired empirical and simulation studies are needed to test model assumptions (Rosenbaum et al. 2024), CKMR represents an unexpectedly promising management tool. Notably, though, CKMR represents just one example of many analyses where power has been improved by more markers (but see Box 1).

Environmental DNA (eDNA) studies have surged in popularity following the genomics revolution (Shokralla et al. 2012; Taberlet et al. 2012; Thomsen and Willerslev 2015). High‐throughput sequencing enables the identification of multiple species simultaneously from a single environmental sample (i.e., eDNA metabarcoding), thereby opening the door to more efficient biodiversity surveys (Chavez et al. 2021) and more cost‐effective methods for tracking species in traditionally undersampled habitats (e.g., Willerslev et al. 1999; Pietramellara et al. 2009). While salmon have been the subject of eDNA studies, most of these are focussed on species detection or quantification (e.g., Baetscher et al. 2024; Spence et al. 2021; Tillotson et al. 2018) rather than population genetics or conservation genomics. Although eDNA studies may be motivated by conservation goals (e.g., Shelton et al. 2019; Wood et al. 2021), only a small number—none of which target salmonids—have begun to explore using eDNA for population genetics based on mtDNA haplotype data (Parsons et al. 2018; Parsons et al. 2024; Sigsgaard et al. 2017). The key limitation for using eDNA from water samples for population genetics is that the number of individual fish contributing to a water sample is typically unknown, requiring some simplifying assumptions about the data set to generate haplotype frequencies (Adams et al. 2019). Yet innovations in data derived from environmental samples—including eRNA (Parsley and Goldberg 2024) and shotgun metagenomics (Garlapati et al. 2019)—suggest that the field of eDNA will continue to progress towards generating population genetic data, with future potential for conservation genomic applications. Additional applications of methods made popular by eDNA (i.e., metabarcoding, metagenomics) include ancient DNA (Oosting et al. 2019), diet analysis (Roy and Boulding 2024), microbiome characterisation (Bozzi et al. 2021) and pathogen detection and monitoring (Peters et al. 2018), which are benefiting salmonid researchers as well as many other studies of nonmodel organisms.

Functional genomics approaches (i.e., GWAS and QTL mapping) have steadily become more precise as a result of high‐density data (Vallejo et al. 2024), but targeted editing of specific genic sequences was prohibitively complex for nonmodel systems before Barrangou et al. (2007) demonstrated that prokaryotic clustered regularly interspaced short palindromic repeats (CRISPR) could be harnessed to cleave DNA at targeted locations and subsequently change amino acid sequences with high precision and accuracy. This unprecedented advancement is still relatively untested in relation to its applications for nonmodel aqua/agricultural and wild species (Winther 2024; but see Kleppe et al. 2022; Phelps et al. 2020). However, both captive and wild populations of salmonids have been subjects of successful CRISPR‐based experiments, which have further deepened our understanding of functional genomics and enabled novel approaches to rapidly identify species (Edvardsen et al. 2014; McKelvey et al. 2016; Datsomor et al. 2019; Williams et al. 2021; Baerwald et al. 2023).

6. Unsolved Needs, Updated Predictions and Emerging Innovations

As we reflect on past predictions and current achievements spurred by the genomics era, we conclude by identifying current needs and major unresolved issues and positing updated predictions for the coming decade of research (Figure 3D). The capacity of genomics to advance research for nonmodel species remains promising. For salmonids as a phylogenetic group, the genomics revolution may have elevated several well‐studied species to ‘model species’ status, with cascading benefits for research of lesser‐studied species. Yet, the pace of genomics research continues to accelerate, and conservation and management applications appear to have lagged behind. While many examples exist for how genomics have informed management decisions (Waples et al. 2020), policy actions are often slowed by resource limitations or bureaucratic processes. We predict the coming decades will see far more management and conservation applications, as managers catch up to the current state of genomics research.

Although computational resources are steadily increasing in availability, modern sequencing efforts still require sometimes costly DNA preparation. Sequencing costs aside (Figure 1), high expenses for proprietary kits, laboratory and bioinformatics personnel, protocol development and equipment continue to make high‐throughput genomic sequencing inaccessible to many researchers, especially in resource‐limited settings (Wasswa et al. 2022; Linck and Cadena 2024; Hogg 2024). Current benchtop technologies, such as MinION nanopore sequencing, are limited in utility to species with small genome sizes or studies with few individuals, as preparation and sequencing of individual samples can take hours (Jain et al. 2018). However, these in‐house sequencers foreshadow a future where researchers can locally prepare and analyse hundreds or thousands of samples within minutes, circumventing the need to outsource to production‐scale sequencing centres. Such innovations will facilitate large‐scale, long‐read WGS, drastically decrease the lag between sample collection and data availability, and reduce the need for targeted genotyping of diagnostic loci; although, targeted loci will likely remain useful for applications where more loci are unnecessary (i.e., ‘genomics‐enabled‐genetics’, Box 1). Whereas current genomics research primarily focusses on SNPs identified via short‐read sequencing, individual‐level genome assemblies generated from accurate long‐read, high‐depth data will open doors to comparative genomic examinations (Dimos and Phelps 2023) and previously overlooked structural (Mérot et al. 2023; Wellenreuther et al. 2019; Harris et al. 2024) and pangenomic variation (Stenløkk 2023; Lecomte et al. 2024). We anticipate the dawn of a pangenomics era, characterised by more comprehensive intraspecific insights into evolutionary mechanisms and associated genomic architectures.

Similarly, multi‐omics approaches are advancing our understanding of biological systems (i.e., Table 1, Prediction 7); however, their relevance and application in conservation and management remains nascent. We foresee the development of G2P workflows that seamlessly combine multi‐omic data types to elucidate adaptive mechanisms and functional pathways, including the burgeoning importance of epigenetic variation (Christensen et al. 2021; Koch et al. 2023). Although there has been some progress in nonmodel species gene annotation in aquaculture relevant contexts (Johnston et al. 2024), lack of ecologically relevant functional annotation remains a challenge for interpreting multi‐omics data from an ecological and evolutionary perspective. Functional databases, such as the Gene Ontology Knowledgebase (Aleksander et al. 2023), will continue to play an important role enabling cross‐species interpretation of the molecular function, biological role and cellular location of gene products (Primmer et al. 2013). However, earlier calls for systematic reporting and description of ecologically relevant gene annotations, such as gene expression responses to ecologically relevant stimuli (Pavey et al. 2012; Macqueen et al. 2017) have not been heeded. Similarly, gene editing technologies (discussed above) will likely play an increasingly important role in G2P research and subsequently in management applications (Phelps et al. 2020). However, while gene editing holds promise for enhancing traits such as disease resistance and climate tolerance in wild and captive populations, it raises concerns about potential adverse impacts on natural diversity and human health. Ethical considerations around synthetic biology in both wild and captive settings must be addressed, and public support will be vital for widespread adoption (Kohl et al. 2019; Redford et al. 2019; Blix and Myhr 2023; Robinson et al. 2024).

Lastly, the integration of artificial intelligence (AI) into genomics workflows is ushering in a new era of progress for nonmodel species research (van Oosterhout 2024). Machine learning (ML) algorithms are already widely used in various genomic applications (Brieuc et al. 2018; Esposito et al. 2019; Cordier et al. 2019; Caudai et al. 2021; Maqsood et al. 2024). Looking forward, integrated AI analysis pipelines are set to revolutionise the accessibility of bioinformatics and the computational processing of genomic data (Miao et al. 2024). Currently, analysing genomic data sets requires highly skilled practitioners, and much effort has been directed toward developing user‐friendly, open‐source bioinformatics tools. Specialised large language models (LLMs) trained on all available ‘omic analysis pipelines’ will dramatically reduce the time needed to learn, modify and apply new analyses. Consistent with the rapid integration of AI, ML and LLMs into many other fields, extreme caution must be exercised in their application and validation (Whalen et al. 2022). Nevertheless, just as genomics has revolutionised genetics applications, the AI revolution promises to lead the next paradigm shift for genomics.

Author Contributions

Samuel A. May led the study and wrote the paper. Samuel W. Rosenbaum wrote the paper and generated figures. Devon E. Pearse edited the paper. Marty Kardos edited the paper. Craig R. Primmer edited the paper. Diana S. Baetscher edited the paper and Robin S. Waples conceptualised the study and wrote the paper.

Disclosure

Benefit‐sharing statement: A research collaboration was developed with scientists in attendance at the 2024 International Conference on Integrative Salmonid Biology, after which additional collaborators were invited to contribute. All collaborators are included as co‐authors. The results of the research have been shared with the broader scientific community and internally reviewed by the home agencies of contributing scientists. This group is committed to continuing to foster international partnerships with a diverse community of scientists and stakeholders.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

All data sources are provided or referenced in the manuscript.