A whole-genome resequencing study investigating 98 mandarin fish has uncovered 32 genome-wide significant single-nucleotide polymorphism associations tied to body dimensions and shape, highlighting specific candidate genes involved in cell survival, immune regulation, and lipid signaling. Mandarin fish, scientifically classified as Siniperca chuatsi, represents China’s most economically valuable freshwater aquaculture species, generating an annual farm output value surpassing four billion dollars across approximately 480,000 metric tons produced in 2023. Despite this massive commercial footprint, selective breeding in the species has historically depended on classical phenotypic mass selection. This traditional approach faces substantial hurdles due to lengthy generation intervals, comparatively modest heritability under farm conditions, and an absence of systematic pedigree tracking. By pairing high-coverage genomic sequencing with advanced mixed-model association methods, researchers sought to establish the first comprehensive map of multi-trait genetic architecture governing physical growth in this prized teleost.
To establish an ultra-high-resolution physical coordinate framework for variant calling, the research team constructed a de novo, telomere-to-telomere reference genome for Siniperca chuatsi. Assembled by combining 59.76 gigabases of ultra-long Oxford Nanopore reads, 200.79 gigabases of PacBio high-fidelity reads, and 223.19 gigabases of chromosome conformation capture data, the resulting gap-free sequence totaled 736.12 megabases across 24 chromosomes with a BUSCO completeness score of 98.6 percent. This telomere-to-telomere resource offered a marked technical improvement over the earlier 758.78-megabase draft assembly, successfully resolving previously inaccessible repetitive elements and telomeric regions. For the association cohort, genomic DNA extracted from fin-clip tissue of 98 fish raised together in a uniform common pond was sequenced on the Illumina NovaSeq 6000 platform to an average depth exceeding 10-fold per individual. Alignment against the gap-free assembly and rigorous GATK quality filtering retained 1,310,398 high-confidence bi-allelic single-nucleotide polymorphisms distributed across the 24 autosomes.
Before conducting genotype-phenotype scans, investigators evaluated the population genetic structure of the cohort using linkage-disequilibrium-pruned markers across three complementary methodologies. Principal component analysis revealed that the first two genomic axes captured 53.31 percent of total genetic variance, partitioning the individuals into four distinct clusters alongside a transitional intermediate cohort. An unrooted neighbor-joining phylogenetic tree reconstructed from identity-by-state distance matrices confirmed this topological division, showing four well-supported clades linked by extended internal branches. Model-based ancestry estimation via ADMIXTURE corroborated this grouping, with cross-validation error reaching an optimal global minimum of 0.257 at six ancestral components. These convergent analyses established that the cultured stock was not a panmictic unit but comprised four primary lineages separating into as many as six finer-scale ancestry groups, displaying clear signatures of historical or commercial admixture that had to be rigorously controlled to avoid spurious statistical associations.
Four primary morphometric characteristics were measured directly from each experimental specimen using standardized ichthyological procedures, namely standard body length, maximum dorso-ventral body height, total body weight, and lateral body thickness. Body length averaged 155.10 millimeters, body height averaged 53.91 millimeters, body weight averaged 91.61 grams, and body thickness averaged 20.55 millimeters. Phenotypic coefficients of variation varied widely from 20.65 percent for length up to 63.85 percent for weight. Pairwise correlations among all four raw linear and mass dimensions were extraordinarily strong, spanning Pearson correlation coefficients between 0.971 and 0.991. Because raw values displayed bimodal distributions that directly mirrored the sub-population genetic clusters, every trait was transformed using rank-based inverse normal transformation to restore statistical normality. Phenotypic principal component analysis of the standardized matrix demonstrated that a single dominant axis, termed Growth-PC1, accounted for 98.33 percent of all phenotypic variance, with factor loadings exceeding 0.99 for all four measurements.
To evaluate the genetic architecture of growth, the researchers deployed both univariate and multivariate linear mixed models implemented in the GEMMA software package. In single-trait evaluations, body height, body weight, and a derived length-to-thickness ratio served as distinct response variables, whereas body length and thickness were channeled into derived proportions and composite indices rather than standalone models. To capture pleiotropic loci affecting multiple dimensions simultaneously, two multivariate combinations were evaluated: a Size-by-Shape model combining Growth-PC1, length-to-height ratio, and Fulton’s condition factor, and a Shape3 model combining Fulton’s condition factor, length-to-height, and length-to-thickness. In all configurations, polygenic relatedness was modeled through a centered genomic relationship matrix, while population stratification was controlled using between two and three genotype principal components. Genomic inflation factors remained strictly well-calibrated between 0.95 and 1.08, confirming that the applied statistical controls adequately eliminated background stratification bias.
Across the five distinct GWAS scans, a strict Bonferroni-corrected threshold of 3.82 times 10 to the negative eighth power identified 32 genome-wide significant single-nucleotide polymorphisms, alongside dozens of suggestive signals reaching a secondary significance threshold of 7.63 times 10 to the negative seventh power. Among single-trait analyses, body weight yielded ten significant variants, body height yielded two, and the length-to-thickness ratio yielded one. The single most significant individual association discovered in the entire study emerged on chromosome 13 at position 10,644,580 within an intronic region of the MRC1 gene, reaching an extraordinary P-value of 5.73 times 10 to the negative fourteenth power in the multivariate Size-by-Shape model. The same MRC1 locus also showed genome-wide significance in the single-trait shape index and the multi-trait Shape3 analysis, demonstrating an exceptionally consistent statistical influence across distinct morphological criteria.
The most uniformly detected cross-trait locus was an intronic single-nucleotide polymorphism located on chromosome 10 at coordinate 3,148,652 inside the BCL2L1 gene, maintaining a minor allele frequency of 0.224. BCL2L1 encodes Bcl-2-like protein 1, a recognized anti-apoptotic regulator that promotes cell survival by antagonizing pro-apoptotic effectors including BAX and BAK. In teleost physiology, the coordinated balance between cellular proliferation and programmed apoptosis governs skeletal muscle hyperplasia and muscle fiber hypertrophy, the direct biological drivers of somatic weight gain. Furthermore, Bcl-xL functions at the interface of cell-death regulation and cellular metabolism, positioning it as a plausible mechanistic bridge between apoptotic control and the metabolic demands of somatic growth.
The biological significance of immune pathways in aquaculture growth was further underscored by findings surrounding MRC1 and Ripk2. Macrophages actively regulate vertebrate satellite cell activation during muscle remodeling, while tissue-resident populations mediate energy metabolism and adipose tissue homeostasis. The study’s lead single-trait marker for both body height and body weight was an intronic variant on chromosome 13 at position 20,451,993 within the Ripk2 gene, which reached a P-value of 1.78 times 10 to the negative tenth power for weight. Ripk2 encodes receptor-interacting serine/threonine-protein kinase 2, an essential signaling transducer operating downstream of NOD receptors to drive inflammatory cytokine release. The detection of two immune-related genes among top growth loci reinforces the concept of an immune–growth trade-off, although researchers noted that determining the mechanistic basis requires further research.
Beyond single-trait discoveries, multivariate linear mixed modeling demonstrated clear statistical advantages by uncovering cryptic loci that completely evaded univariate detection. Specifically, a discrete genomic window spanning 3.66 to 3.73 megabases on chromosome 18 emerged solely within the Size-by-Shape multivariate model, containing seven significant markers across the PITPNM2 and Zswim6 genes. Because individual univariate scans were underpowered to register the modest, correlated contributions of this region to separate traits, joint modeling was required to pool statistical signal across correlated phenotypes. Within this multivariate-specific locus, functional annotation using ANNOVAR and PolyPhen-2 highlighted a nonsynonymous coding substitution in PITPNM2 at chromosome 18, position 3,624,313. This missense variant alters codon 1254 from arginine to histidine, scoring a HumDiv prediction of 0.997 and a HumVar score of 0.949, marking it as a prioritized computational candidate whose substitution could directly alter protein function.
Additional cross-trait loci illuminated other metabolic and transport pathways contributing to body conformation in mandarin fish. ABCG1 encodes an ATP-binding cassette transporter central to cellular oxysterol and cholesterol export, maintaining membrane lipid composition and modulating lipid deposition in teleost tissues. Another locus on chromosome 5 tagged FBXL17, an F-box protein-coding gene, while an intergenic signal on chromosome 4 was mapped adjacent to CHRM4 and Dgkz, broadening the spectrum of physiological systems implicated in mandarin fish morphogenesis.
Using genomic restricted maximum likelihood estimation in GCTA, researchers calculated SNP-based heritability estimates based on 1,241,008 autosomal markers across the 98 fish. Body height yielded a heritability estimate of 0.527 with a standard error of 0.287, narrowly missing conventional statistical significance at a P-value of 0.060. Body weight exhibited an estimated heritability of 0.369 with a standard error of 0.296, while the length-to-thickness ratio registered 0.330 with a standard error of 0.325. A post hoc statistical power calculation revealed that with a sample cohort of 98 individuals, 80 percent power at the Bonferroni threshold required a minimum phenotypic variance explained of approximately 29.1 percent. Consequently, the observed lead variants, which accounted for 28.6 to 36.7 percent of phenotypic variance, represent the upper tier of large-effect loci, whereas polygenic variants of small to moderate effect remained beneath the detection limit.
The authors highlighted several critical caveats when interpreting these genomic findings for breeding applications. Foremost, the discovery cohort consisted of 98 individuals, meaning that detected effect sizes are susceptible to statistical overestimation, and the study was inherently powered only for major-effect loci. Furthermore, because body dimensions share near-complete phenotypic correlations exceeding 0.97, the repeated discovery of markers across multiple models represents shared phenotypic signal rather than independent biological validation. Strong linkage disequilibrium between lineages, driven by the four ancestral groups, also means some lead variants may tag ancestral haplotype blocks rather than isolated functional mutations. In addition, fish ages and individual sexes were unrecorded, leaving potential sex-linked growth dimorphism uncontrolled. Consequently, researchers emphasized that the prioritized computational candidates, including the PITPNM2 missense alteration, require rigorous functional knockdown experiments, gene expression analyses, and independent validation across diverse breeding populations before integration into commercial marker-assisted selection.
Subject of Research: Genome-wide association study of growth and morphometric traits in mandarin fish
Article Title: Genome-wide association study of growth-related traits in mandarin fish (Siniperca chuatsi): identification of cross-trait associated loci and candidate genes using single-trait and multivariate linear mixed models
Article References: Yang, T., Hanif, Q., Yao, X., Cheng, J., Shehzad, S., Zhang, X., & Cui, R. (2026). Genome-wide association study of growth-related traits in mandarin fish (Siniperca chuatsi): identification of cross-trait associated loci and candidate genes using single-trait and multivariate linear mixed models. Frontiers in Genetics, 17, Article 1956014. https://doi.org/10.3389/fgene.2026.1956014
Image Credits: AI Generated
DOI: 10.3389/fgene.2026.1956014
Keywords: Siniperca chuatsi, mandarin fish, aquaculture genomics, genome-wide association study, multivariate linear mixed models, quantitative trait loci, BCL2L1, MRC1, PITPNM2, Ripk2, body weight, body height, morphometrics, population structure, telomere-to-telomere assembly, single nucleotide polymorphisms, marker-assisted selection, genomic selection
News Source: Juliet Wilcox. (November 14, 2026). Mandarin Fish Genome Study Identifies Key Genetic Loci Regulating Growth Traits. Scienmag.



