# Forward–reverse validation of dwarf coconut domestication through ABC1K7 **Authors:** Ninghuan You¹²†, Wenrao Li²†, John Martin¹, Jing Chen³, Huangsheng Ling³, Mengxing Huang⁴, Yu Zhang⁵, Hongxing Cao¹*, Chengxu Sun¹* ¹ Coconut Research Institute, Chinese Academy of Tropical Agricultural Sciences, Wenchang, Hainan 571339, China ² School of Life Sciences, Henan University, Kaifeng, Henan 475004, China ³ Haikou Affiliated Hospital of Central South University Xiangya School of Medicine, Haikou, Hainan 570208, China ⁴ College of Information and Communication Engineering, Hainan University, Haikou, Hainan 570228, China ⁵ College of Computer Science and Technology, Hainan University, Haikou, Hainan 570228, China † These authors contributed equally to this work * Corresponding authors: suncx@catas.cn (C. Sun); caohx@catas.cn (H. Cao) --- **Running title:** Forward–reverse validation of coconut domestication **Keywords:** Forward–reverse breeding; ABC1K7; selective sweep; dwarf domestication; coconut (*Cocos nucifera*); multi-omics --- ## Abstract Perennial tree crop breeding is slow: traditional germplasm evaluation requires dozens of indicators measured over several years. Breeders routinely simplify these criteria without knowing whether the retained indicators track genuine biological signals or correlated noise. We reconstructed a 17-year coconut dwarf breeding program (2009–2026) and examined each empirical decision against multi-omics data. The forward path compressed 36 evaluation indicators to 5 (86% reduction), producing two nationally registered sweet water varieties, Wenye 5 (W5) and Wenye 6 (W6). The reverse path integrated whole-genome resequencing (135 individuals), metabolomics (624 metabolites), transcriptomics (27 samples), CT phenomics (7,686 slices), and proteomics (7,550 proteins). All data paths converged on ABC1K7: a Y→F missense mutation in the kinase domain (Chr5:20,238,027; −log₁₀P=39.66; π ratio peak=28.7). This mutation is associated with disrupted flavonoid pathway canalization (CV=0.172 in CK vs. 0.485 in W5), linked to a shift in carbon allocation from defense to growth metabolism (35.1-fold increase in total amino acid peak area at 2 months). The 5-indicator system captured key traits associated with this mutation, carried on a 600-kb low-recombination haplotype block that swept from 55% to 92% homozygous frequency. When indicator simplification succeeds in perennial breeding, the retained criteria likely capture a major-effect locus. **Key message:** A 17-year breeding program's empirical indicator simplification was examined against multi-omics data; all evidence converged on a single ABC1K7 mutation. --- ## Introduction Coconut (*Cocos nucifera* L.) is a pantropical perennial tree crop cultivated across 86 countries, supporting the livelihoods of millions of smallholder farmers [1]. Sweet water coconut—defined by total soluble solids (TSS) ≥ 6 °Brix in the liquid endosperm—is valued as a fresh beverage [2,17]. China imports over 500,000 tonnes of fresh coconuts annually, reflecting insufficient domestic supply of specialized sweet water varieties (Supplementary Note S1). The bottleneck is not genotype availability but evaluation speed: traditional morphological and quality assessment uses 36 indicators measured over approximately 36 months (Supplementary Note S1). Despite advances in molecular and genomic approaches to coconut improvement [18], systematic acceleration of germplasm evaluation remains challenging. Efforts to accelerate perennial crop breeding through indicator simplification are documented in macadamia [3] and coconut (Supplementary Note S1). Perennial fruit crops pose unique challenges for domestication and breeding because their long generation times preclude the rapid cycles available to annual crops [15]. Whether, when breeders discard 86% of evaluation criteria, they are selecting on valid biological indicators or on correlated noise remains unknown. Retrospective molecular validation can address this, yet such validation is rare in perennial crops because breeding actions and molecular tools are separated by decades. The Hainan coconut breeding program offers an opportunity to examine this question. From germplasm introduction (2003) to variety registration (2017) to molecular decoding (2023–2026), the program produced two nationally registered sweet water varieties, Wenye 5 (W5) and Wenye 6 (W6), and cut evaluation indicators by 86% (Supplementary Note S1). To interrogate these breeding decisions at molecular resolution, we generated five data layers: a 24-SNP variety authentication panel, whole-genome resequencing of 135 individuals, multi-omics profiling of endosperm development, and CT-based phenomics [5] and structural imaging [4]. We examined three questions. Did the indicator simplification produce the intended varieties? Does population genomics identify a molecular target of selection? Can multi-omics account for why the simplified indicators worked? All three converge on ABC1K7 (chromosome 5). --- ## Results ### Forward path: empirical breeding and variety development #### Germplasm foundation and indicator simplification The breeding program began with two waves of Vietnamese coconut introduction into Hainan. The first wave (1982–2003) involved private enterprises importing bulk seed nuts; the second wave (2003) was a targeted introduction of approximately 150,000 seed nuts from Bến Tre Province (Supplementary Note S1). Systematic evaluation by CRI-CATAS began in 2009. Between 2009 and 2014, the program collected 20 Vietnamese accessions and established a dual conservation system combining *ex situ* and *in situ* preservation. By 2010, the collection expanded to 135 accessions (Supplementary Note S1). From the initial 20 accessions, four elite types were identified: V15-11 (Bung, red fruit, 102 nuts/plant/year), V15-17 (Dau, green fruit, 112 nuts/plant/year), V15-18 (EO, red-brown fruit, sweet water, 93 nuts/plant/year), and V15-20 (early-bearing dwarf). V15-17 and V15-18 became the genetic founders of the sweet water breeding program. To speed up germplasm evaluation, we compressed the standard 36-indicator system to 5 core indicators using PCA and hierarchical clustering. The first two principal components cumulatively explained >70% of total variance (KMO > 0.6, Bartlett's test P < 0.001). Indicators with loadings < 0.3 on PC1–PC2 were removed, achieving 86.1% dimensionality reduction (Supplementary Note S1, Figure 5A). #### Variety selection and registration Application of the 5-indicator system to the 135-accession collection identified two superior strains: 15-19 (from V15-18 EO line, red-brown fruit) and 15-17 (from V15-17 Dau line, green fruit). Following four-region field testing, 15-19 was registered as Wenye 5 (W5; certificate no. QIONG R-ETS-001-2017) and 15-17 as Wenye 6 (W6; certificate no. QIONG R-ETS-002-2017), both receiving plant variety rights in 2020 (CNA20181058.5, CNA20181059.4). Key phenotypes include earlier flowering (40 vs. 78 months), higher TSS (7.5–8.2 vs. 5.5–6.0 °Brix), and distinct fruit color (Supplementary Note S1). ### SNP panel validates variety identity We developed a 24-SNP panel from whole-genome resequencing of 310 individuals to authenticate the three variety groups. In a 77-sample blind test, the panel achieved perfect classification (Naive Bayes, balanced accuracy = 1.00, 95% Wilson CI: 95.3–100%). Nineteen of 24 SNPs (79%) clustered within a ~6 Mb region on Chromosome 5 (P = 4.14 × 10⁻¹⁹, Figure 5B). The panel's core finding—Chr5 as the primary differentiation chromosome—serves as the entry point for population genomic analysis. ### Population genomics identifies ABC1K7 as the target of selection Whole-genome resequencing of 135 individuals (45 W5, 19 W6, 71 CK) at ~15× coverage identified the core differentiation interval (CDI) on Chromosome 5: a ~6 Mb region (Chr5:16.2–22.1 Mb) characterized by elevated Fst, reduced nucleotide diversity (π) in W5, and negative Tajima's D. Together, these three statistics constitute a standard signature of recent positive selection [14] (Figure 1, Figure 3C). Within the CDI, we fine-mapped a ~600-kb window (Chr5:19.75–20.35 Mb) where all three statistics converge. The π ratio (CK/W5) reached 28.7 at position 19.98 Mb—approximately 17-fold above the chromosome 5 median (π ratio = 1.7). Genome-wide window-averaged Fst = 0.014; within the CDI, per-SNP Fst reached 0.35. Tajima's D in the 100-kb window containing the ABC1K7 GWAS peak (window start 20.18 Mb, spanning 20.18–20.28 Mb) was −2.20 (N = 391 SNPs), with the most negative values in the CDI reaching −2.39 at 17.26 Mb (Supplementary Data S1). In the immediately adjacent downstream windows (20.24–20.35 Mb), Tajima's D shifted to +2.13 to +2.15, consistent with post-sweep recovery of nucleotide diversity. Across the full CDI, Tajima's D ranged from −2.39 to +5.45 in 100-kb windows, with negative values concentrated in the sweep-affected regions. Among 11 annotated genes in the CDI, ABC1K7 (COCNU_05G012990) harbors a non-synonymous mutation differentiating W5 from CK: a Y→F substitution at a residue conserved across the plant ABC1K family. Of the 11 CDI genes screened for coding-sequence variants, this was the only non-synonymous change detected. GWAS independently identified ABC1K7 as the strongest signal for fruit peel color a* (−log₁₀P = 39.66 at Chr5:20,238,027; β = −0.70), explaining 68% of phenotypic variance. The derived F allele was homozygous in 92% of W5 individuals and 55% of CK individuals; it was absent from W6 (0%). ROH analysis confirmed that 76% of W5, 59% of CK, and 37% of W6 individuals carry the CDI haplotype in the homozygous state (Figure 3A, 3B). ### Multi-omics evidence for defense release Widely targeted metabolomic profiling (624 metabolites, 12 compound classes) across six developmental stages showed consistent divergence. CK accumulated flavonoid glycosides as the dominant class, with isorhamnetin-3-O-rutinoside 17.7-fold higher than W5—consistent with the central role of the flavonoid biosynthetic pathway in plant secondary metabolism [22]. W5 showed substantial enrichment of nucleotide and amino acid metabolites, with dIMP showing the largest fold-change (34.6-fold). This flavonoid-to-amino-acid ratio inversion marks the metabolic signature of defense release. Transcriptomic analysis of 27 samples focused on 34 core flavonoid biosynthetic enzymes. The coefficient of variation (CV) in gene expression was used to quantify pathway coordination, following the concept of canalization—the tendency of developmental pathways to produce consistent phenotypes despite genetic and environmental variation [16]. Simulations of gene regulatory networks predict that domestication relaxes canalization, generating the elevated phenotypic variance observed in domesticated forms [21]. In CK, expression was tightly coordinated (mean CV = 0.172); in W5, coordination weakened considerably (mean CV = 0.485). The CV gradient was progressive: CK (0.172) < W6 (0.218) < W5 (0.485). The decanalization was specific to the flavonoid pathway: housekeeping genes showed comparable CV values (0.08–0.18) across all varieties. This pattern mirrors the metabolomic decanalization of flavonoid accumulation reported in coconut water [7]. Quantitative proteomics comparing W5 and CK at the mid-developmental stage (5–7 months) identified 2,865 differentially abundant proteins (1,695 up, 1,170 down; fold-change > 1.5, FDR < 0.05). Chloroplast-associated proteins were enriched among those with higher abundance in W5, consistent with plastid localization of ABC1K family kinases [8] and chloroplast-to-nucleus retrograde signaling [9]. CT imaging of mature coconuts during germination has been used to characterize internal structural changes [4]. Here, CT-based phenotyping revealed accelerated tissue differentiation in W5. Shell density increased from 0 to +202 HU in W5 vs. +256 HU in CK. Endosperm density rose from 0 to +37 HU. W5 reached developmental milestones approximately 2 months earlier than CK (Figure 2). ### Reverse path: breeding verification Four lines of evidence indicate that the 5-indicator system captured phenotypic footprints of the ABC1K7 mutation. **TSS as a terminal integrator.** W5 and W6 maintained TSS of 7.5–8.2 °Brix versus 5.5–6.0 for CK (Supplementary Note S1). At 2 months, total amino acid peak area was 4,911,374 in W5 versus 139,991 in CK, a 35.1-fold enrichment that reflects the carbon reallocation from defense to growth metabolism, consistent with the observed TSS elevation. **Fruit color as a visible marker.** The red-brown pericarp that identified V15-18 as an elite type is a direct phenotypic readout of ABC1K7, confirmed by GWAS (−log₁₀P = 39.66). **Developmental timing.** The earliness indicator (40 months to first flowering vs. 78 months) reflects accelerated development validated by CT imaging. **The CDI haplotype as a molecular tag.** The derived ABC1K7 allele, carried in the homozygous state by 76% of W5 individuals (ROH confirmed), serves as a molecular proxy for variety performance. Breeders selected for the phenotype (sweet water + red-brown fruit + early bearing); population genomics shows these phenotypes are coupled through a single ~600-kb haplotype block. --- ## Discussion ### The forward–reverse framework The 36→5 indicator simplification tracked genuine biological signal, not correlated noise. Two of the 5 indicators are associated with ABC1K7: TSS (via metabolic reallocation from defense to growth) and fruit color (GWAS, −log₁₀P=39.66). The remaining three (fruit weight, pedicel-stigma distance, fat/acid ratio) are correlated traits; their mechanistic connection to ABC1K7 remains to be characterized. The forward path identified the phenotypes, the reverse path identified the genotype, and both converged on the same locus. This forward–reverse logic builds on genetic approaches where forward screens identify phenotypes and reverse analysis reveals the underlying molecular basis [19]. Can accelerated screening produce genuinely superior varieties in perennials? In this case, the accelerated protocol converged on the primary domestication locus within 17 years—a timeline comparable to annual crop improvement (Figure 4). ### ABC1K7 as a domestication accelerator The Y→F substitution affects a residue conserved across the plant ABC1K family [8] and is predicted to impair kinase function. ABC1K kinases contribute to chloroplast function; the conserved active-site aspartate of ABC1K1 is required for kinase activity [23], and loss of specific family members has been associated with altered photosynthetic performance in Arabidopsis [8]. The observed changes in W5 suggest a viable trade-off: reduced flavonoids and increased amino acids are consistent with a shift from defense to growth metabolism, though direct functional evidence is pending. The mutation resides within a 600-kb low-recombination block where all three standard sweep statistics—π reduction (28.7-fold), Fst elevation (0.35), and negative Tajima's D (−2.20)—converge at ABC1K7. Combined with coconut's mating system, the allele frequency difference between CK (55%) and W5 (92%) is consistent with strong selection at this locus. SSR marker analysis independently supports a common origin of self-pollinating dwarf coconut in Southeast Asia under domestication [13]. ### Growth–defense trade-off magnitude Growth–defense trade-offs, driven largely by hormonal crosstalk rather than direct metabolic expenditure, represent a fundamental constraint on plant productivity [10,20]. The 17.7-fold flavonoid reduction paired with a 34.6-fold increase in the nucleotide dIMP reflects a marked metabolic shift. Plants mitigate such costs primarily through restricted, inducible expression of resistance [11]. The larger dynamic range in coconut may reflect the higher baseline flavonoid investment required for long-lived perennial tissues. ### Limitations The temporal resolution limits statistical power for time-series comparisons across developmental stages. Proteomic comparison reported here is between W5 and CK at the mid-developmental stage; W5–W6 proteomic comparison was performed at a single time point and is described in a companion study. Functional validation of ABC1K7 through CRISPR is ongoing. ### Broader application The forward–reverse framework applies to any perennial breeding program with documented evaluation protocols, archived germplasm, and multi-omics capability. As the contrasting population genetic dynamics of annual and perennial domestication become better characterized [12,24], the framework can identify the molecular basis of empirical breeding success. --- ## Materials and Methods ### Germplasm and phenotype data Breeding program data were extracted from CRI-CATAS institutional records (Supplementary Note S1). Variety registration timelines were verified against official certificates (CNA20181058.5; CNA20181059.4). ### SNP panel development The 24-SNP panel was developed from whole-genome resequencing of 310 individuals, using Fst-based pre-screening and Naive Bayes classification. ### Whole-genome resequencing and population genetics DNA from 135 individuals was sequenced on DNBSEQ-T7 to ~15× coverage. Reads were aligned to the coconut reference genome [6] using BWA-MEM v0.7.17. Population genetic parameters were computed with vcftools v0.1.16 using 100-kb windows. GWAS used GEMMA v0.98 (MLM, P+K). Significance threshold: −log₁₀(P) > 6.0. ROHs were identified with PLINK v1.9. Tajima's D values were computed with vcftools v0.1.16 using 100-kb windows (Supplementary Data S1). ### Metabolomics Widely targeted metabolomics (624 metabolites) was performed on coconut water across six developmental stages with three biological replicates. UPLC-ESI-MS/MS used a Shimadzu Nexera X2 with AB Sciex QTRAP 6500+. Metabolomic methods followed the approach described in [7]. ### Transcriptomics RNA from 27 samples (3 varieties × 3 stages × 3 replicates) was sequenced on NovaSeq 6000. Reads were processed with fastp v0.20.0, aligned with HISAT2 v2.2.1, quantified with StringTie v2.1.5. Differential expression used edgeR v3.32.1. ### Proteomics TMT-labeled quantitative proteomics was performed on liquid endosperm. Proteins were extracted, digested with trypsin, and analyzed on an Orbitrap Exploris 480 mass spectrometer. Differential abundance was tested using two-sided Student's t-test with fold-change > 1.5 with FDR < 0.05. Raw data are deposited under iProX accession IPX0017960000; independently processed data used for W5–W6 comparison are described in a companion study. ### CT phenomics DICOM images (7,686 slices) were acquired on a SOMATOM Definition AS+ [4]. Segmentation used DeepLabV3+ (Dice = 0.93) following the approach of [5]. ### Statistical analysis PCA was performed in R v4.2. Metabolomics differential analyses used Welch's t-test with BH correction (q < 0.05), comparing W5 vs CK at each developmental stage. --- ## Acknowledgments We thank the coconut breeding team at CRI-CATAS for maintaining the germplasm collection and field trial network. --- ## Author Contributions (CRediT) **Conceptualization:** C.S., H.C. **Data curation:** N.Y., W.L. **Formal analysis:** N.Y., W.L., J.M., J.C., H.L., M.H., Y.Z. **Funding acquisition:** C.S., H.C. **Investigation:** N.Y., W.L., J.M., J.C., H.L., M.H., Y.Z. **Methodology:** C.S. **Project administration:** C.S., H.C. **Supervision:** C.S., H.C. **Validation:** N.Y., W.L. **Visualization:** J.M., M.H. **Writing – original draft:** C.S. **Writing – review & editing:** C.S., N.Y., W.L., J.M. --- ## Declarations **Competing interests:** The authors declare that they have no competing interests. **Funding:** This work was supported by the Key Research and Development Project of Hainan Provincial Department of Science and Technology (ZDYF2026XDNY143), the Central Finance Forestry Science and Technology Promotion Demonstration Fund Project of Hainan Province (QIONG〔2024〕TG07), and the International Science and Technology Cooperation Research and Development Project of Hainan Provincial Department of Science and Technology (GHYF2025027). --- ## Data Availability Whole-genome resequencing data have been deposited in the NCBI Sequence Read Archive under BioProject accession PRJNA1477953. Proteomics and metabolomics datasets are deposited under iProX project IPX0017960000 as separate sub-datasets. Metabolomics data are also provided as Supplementary Data. Tajima's D values across the CDI are provided as Supplementary Data S1. --- ## References [1] Perera L, Perera SACN, Bandaranayake CK, et al. Coconut. In: Vollmann J, Rajcan I, editors. Oil Crops. New York: Springer; 2009. p. 369–396. DOI: 10.1007/978-0-387-77594-4_12. [2] Prades A, Dornier M, Diop N, Pain JP. Coconut water uses, composition and properties: a review. Fruits. 2012;67(2):87–107. DOI: 10.1051/fruits/2012002. [3] Topp B, Nock C, Hardner C, Alam M. Macadamia (*Macadamia* spp.) breeding. In: Al-Khayri JM, Jain SM, Johnson DV, editors. Advances in Plant Breeding Strategies: Nut and Beverage Crops. Cham: Springer; 2019. p. 221–251. DOI: 10.1007/978-3-030-23112-5_7. [4] Lin S, Sun C, Luo L, Huang M, Cao H, Hu J, et al. The observation of internal structure changes and survival prediction modeling of mature coconut during germination based on computed tomography imaging. Industrial Crops and Products. 2025;233:121396. DOI: 10.1016/j.indcrop.2025.121396. [5] Liu Q, Zhang Y, Chen J, Sun C, Huang M, Che M, et al. An improved Deeplab V3+ network based coconut CT image segmentation method. Frontiers in Plant Science. 2023;14:1139666. DOI: 10.3389/fpls.2023.1139666. [6] Xiao Y, Xu P, Fan H, et al. The genome draft of coconut (*Cocos nucifera*). GigaScience. 2017;6(11):1–11. DOI: 10.1093/gigascience/gix095. [7] Hou M, John Martin JJ, Song Y, Wang Q, Cao H, Li W, et al. Dynamics of flavonoid metabolites in coconut water based on metabolomics perspective. Frontiers in Plant Science. 2024;15:1468858. DOI: 10.3389/fpls.2024.1468858. [8] Lundquist PK, Davis JI, van Wijk KJ. ABC1K atypical kinases in plants: filling the organellar kinase void. Trends in Plant Science. 2012;17(9):546–555. DOI: 10.1016/j.tplants.2012.05.010. [9] Chi W, Feng P, Ma J, Zhang L. Metabolites and chloroplast retrograde signaling. Current Opinion in Plant Biology. 2015;25:32–38. DOI: 10.1016/j.pbi.2015.04.006. [10] Züst T, Agrawal AA. Trade-offs between plant growth and defense against insect herbivory: an emerging mechanistic synthesis. Annual Review of Plant Biology. 2017;68:513–534. DOI: 10.1146/annurev-arplant-042916-040856. [11] Karasov TL, Chae E, Herman JJ, Bergelson J. Mechanisms to mitigate the trade-off between growth and defense. Plant Cell. 2017;29(4):666–680. DOI: 10.1105/tpc.16.00931. [12] Gaut BS, Díez CM, Morrell PL. Genomics and the contrasting dynamics of annual and perennial domestication. Trends in Genetics. 2015;31(12):709–719. DOI: 10.1016/j.tig.2015.10.002. [13] Perera L, Baudouin L, Mackay I. SSR markers indicate a common origin of self-pollinating dwarf coconut in South-East Asia under domestication. Scientia Horticulturae. 2016;211:255–262. DOI: 10.1016/j.scienta.2016.08.028. [14] Nielsen R, Williamson S, Kim Y, Hubisz MJ, Clark AG, Bustamante C. Genomic scans for selective sweeps using SNP data. Genome Research. 2005;15(11):1566–1575. DOI: 10.1101/gr.4252305. [15] Miller AJ, Gross BL. From forest to field: perennial fruit crop domestication. American Journal of Botany. 2011;98(9):1389–1414. DOI: 10.3732/ajb.1000522. [16] Waddington CH. Canalization of development and the inheritance of acquired characters. Nature. 1942;150(3811):563–565. DOI: 10.1038/150563a0. [17] Yong JWH, Ge L, Ng YF, Tan SN. The chemical composition and biological properties of coconut (*Cocos nucifera* L.) water. Molecules. 2009;14(12):5144–5164. DOI: 10.3390/molecules14125144. [18] Ramesh SV, Sudha R, Niral V, et al. Enhancing genetic gain in coconut: conventional, molecular, and genomics-based breeding approaches. In: Gosal SS, Wani SH, editors. Accelerated Plant Breeding, Volume 4. Cham: Springer; 2022. p. 323–362. DOI: 10.1007/978-3-030-81107-5_10. [19] Jankowicz-Cieslak J, Till BJ. Forward and reverse genetics in crop breeding. In: Al-Khayri JM, Jain SM, Johnson DV, editors. Advances in Plant Breeding Strategies: Breeding, Biotechnology and Molecular Tools. Cham: Springer; 2015. p. 215–240. DOI: 10.1007/978-3-319-22521-0_8. [20] Huot B, Yao J, Montgomery BL, He SY. Growth–defense tradeoffs in plants: a balancing act to optimize fitness. Molecular Plant. 2014;7(8):1267–1287. DOI: 10.1093/mp/ssu049. [21] Burban E, Tenaillon MI, Le Rouzic A. Gene network simulations provide testable predictions for the molecular domestication syndrome. Genetics. 2022;220(2):iyab214. DOI: 10.1093/genetics/iyab214. [22] Shirley BW. Flavonoid biosynthesis: 'new' functions for an 'old' pathway. Trends in Plant Science. 1996;1(11):377–382. DOI: 10.1016/S1360-1385(96)80312-8. --- [23] Turquand M, Justo Da Silva A, Pralon T, Longoni F. The conserved active site aspartate residue is required for the function of the chloroplast atypical kinase ABC1K1. Frontiers in Plant Science. 2024;15:1491719. DOI: 10.3389/fpls.2024.1491719. [24] Helliwell EE, Scascitelli M, Blackman BK. Genetic analysis of domestication parallels in annual and perennial sunflowers (*Helianthus* spp.): routes to crop development. Frontiers in Plant Science. 2020;11:834. DOI: 10.3389/fpls.2020.00834. ## Figure Legends **Figure 1. Population genomics of the ABC1K7 selective sweep on Chromosome 5.** (A) Tajima's D across the core differentiation interval (CDI, Chr5:16.2–22.1 Mb) in 100-kb windows. Blue bars indicate negative values; red bars indicate positive values. The yellow highlight marks the ABC1K7 window (20.18–20.28 Mb). The dashed line indicates the GWAS peak position (20.24 Mb). (B) GWAS Manhattan plot for fruit peel color a* on Chromosome 5. The red point marks the peak signal at ABC1K7 (−log₁₀P = 39.66). (C) Window-averaged Fst (Weir & Cockerham) between W5 and CK across the CDI. The yellow highlight marks the ABC1K7 window. **Figure 2. Multi-omics evidence for defense release in W5.** (A) Fold changes of key metabolite classes in W5 relative to CK. (B) Coefficient of variation (CV) of 34 flavonoid biosynthetic gene expression levels in CK, W6, and W5, showing progressive decanalization. (C) Fold enrichment of protein categories among differentially abundant proteins in W5. (D) CT-based tissue density measurements (Hounsfield Units) in W5 and CK shell and endosperm. **Figure 3. ABC1K7 allele frequencies and sweep statistics.** (A) Homozygous frequency of the derived ABC1K7 F allele in W5 (92%), CK (55%), and W6 (absent). (B) Proportion of individuals carrying the CDI haplotype in the homozygous state (ROH analysis). (C) Summary of three standard selective sweep statistics at the ABC1K7 locus: π ratio (CK/W5), per-SNP Fst (W5 vs CK), and absolute Tajima's D. **Figure 4. Forward–reverse breeding validation framework.** Upper path: empirical breeding from germplasm introduction (1982–2003) through variety registration (2017–2020), compressing 36 evaluation indicators to 5 core criteria. Lower path: molecular decoding through five data layers (SNP panel, population genomics, metabolomics, transcriptomics, proteomics/CT), all converging on ABC1K7. **Figure 5. Breeding indicator simplification and SNP panel validation.** (A) Reduction from 36 traditional indicators to 5 core indicators (86% reduction) used in the accelerated evaluation system. (B) Performance of the 24-SNP variety authentication panel in a 77-sample blind test (Naive Bayes classifier). --- ## Supplementary Note S1: Breeding program data sources The 17-year breeding program history described in this paper is documented in two Chinese-language institutional reports that are not accessible as standard academic publications: - Coconut Research Institute, CATAS. Hainan Provincial Science and Technology Progress Award nomination document — Breeding and application of Wenye coconut series. 2024. (in Chinese) - Coconut Research Institute, CATAS. National Agriculture, Animal Husbandry and Fishery Harvest Award application document. 2023. (in Chinese) These documents contain the primary data on germplasm introduction history, indicator selection methodology (PCA, KMO, Bartlett's test), variety phenotypes (TSS, flowering time, fruit traits), and variety registration details. Variety registration certificates (W5: QIONG R-ETS-001-2017; W6: QIONG R-ETS-002-2017) were issued by the Hainan Provincial Seed Station. Plant variety rights (CNA20181058.5, CNA20181059.4) were granted in 2020. Data on China's fresh coconut imports are from Chinese customs statistics as cited in the above documents. --- ## Supplementary Data S1: Tajima's D across the CDI Tajima's D values for 100-kb windows across the CDI (Chr5:16.2–22.1 Mb) computed by vcftools v0.1.16. Selected windows: | Window start (Mb) | N_SNPs | Tajima's D | |:--|:--:|:--:| | 16.19 | 311 | 0.47 | | 16.30 | 103 | −1.90 | | 16.31 | 367 | −1.90 | | 16.46 | 259 | −2.07 | | 17.03 | 152 | −2.04 | | 17.04 | 134 | −2.32 | | 17.26 | 183 | −2.39 | | 20.18 | 391 | −2.20 | | 20.24 | 183 | 2.13 | | 20.25 | 142 | 2.15 | | 22.10 | 106 | 0.49 | Full data available upon request. --- *Version: V7.12.2 | Date: 2026-06-30 | Journal: Horticulture Research | Status: internal audit*