# Quantitative evidence for developmental canalization: CT--metabolome integration reveals a growth--defense axis in coconut fruit development

Huanning You¹²†, Huangsheng Ling³†, Jing Chen³, Wenrao Li¹, 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

† Co-first authors \| \* Corresponding authors: Hongxing Cao (caohx@catas.cn), Chengxu Sun (suncx@catas.cn)

**Received:** ｜ **Revised:** ｜ **Accepted:**

**Abstract**

Coconut (*Cocos nucifera* L.) is one of the most important oil crops in the tropics, yet the structural--metabolic differences in fruit development between varieties---and the biological logic behind them---remain poorly understood. Here we integrate CT phenomics (7,686 DICOM slices across five varieties × five developmental stages) with widely targeted LC-MS/MS metabolomics (624 metabolites) to dissect the structural--metabolic divergence in fruit development between the elite dwarf variety W5 (Wenye 5) and the local tall landrace CK (Hainan Lvgao). CT phenotyping revealed that four tissue-specific CT values (Hounsfield Units, HU) constitute a quantitative "developmental clock": fibre layer CT dropped from −13 to −795 HU (mesocarp air-filling), shell CT rose from 0 to +202 HU (mineralisation), endosperm CT rose from 0 to +37 HU (solidification), and coconut water CT remained stable at 19--27 HU. The four-layer density span of approximately 1,200 HU---from near air to near bone density---makes CT the optimal non-destructive modality for this system. Metabolomic profiling uncovered a pronounced metabolic bifurcation between W5 and CK: W5 was enriched in nucleotides, amino acids, and free fatty acids (growth-oriented metabolism), whereas CK accumulated flavonoids (42 of 44 species higher than W5, up to 32.5-fold), tannins (100% enriched), and phenolic acids (defence-oriented metabolism). Systematic correlation analysis of 18 CT parameters against 624 metabolites identified 3,376 significant associations in W5 (FDR q \< 0.05), of which 55 pairs reached \|r\| \> 0.99. By contrast, CK had no single pair exceeding \|r\| \> 0.95. We do not interpret this asymmetry as breeding having "tightened" structural--metabolic coupling; instead, we propose that W5, as a genetically homozygous variety bred over multiple generations, exhibits highly canalized developmental programs---all individuals follow nearly identical trajectories, producing tight statistical correlations at the population level---while CK, as a genetically heterogeneous landrace, shows divergent developmental programs that dilute correlation strength. CT--metabolite association strength may therefore serve as a quantitative indicator of developmental canalization, pointing toward DNA-level validation. A random forest classifier based on shell thickness and embryo size predicted seed germination with a hold-out AUC of 0.716, providing a practical non-destructive screening tool. The contribution of this study is organised into three tiers: at the descriptive level---the first systematic CT phenomic record of coconut fruit development; at the mechanistic level---metabolomic interpretation of CT trajectory chemistry; and at the hypothesis-generating level---proposing CT--metabolite association strength as a statistical signal of developmental canalization, pointing toward genetic causal validation.

**Keywords:** Coconut \| CT phenomics \| Metabolomics \| Growth--defense trade-off \| Developmental canalization \| Multi-omics integration

## 1 \| Introduction

Coconut (*Cocos nucifera* L.) is a cornerstone crop of tropical agriculture, with an annual production exceeding 60 million tons that supports the livelihoods of millions of smallholder farmers across Asia, Africa, and the Pacific (Perera et al., 2009). In Hainan, China---a major coconut-producing region---systematic breeding programmes have developed the Wenye (WY) series of dwarf varieties (Lu and Liu, 2021). These varieties were selected over decades from local tall landraces and exhibit prominent agronomic traits including early maturation, dwarf stature, high yield, and thin shell (Li et al., 2020). However, the structural and biochemical differences in fruit development between these dwarf varieties and local tall landraces have not been systematically characterised.

X-ray computed tomography (CT) has emerged as a powerful non-destructive phenotyping tool in plant science. CT values (Hounsfield Units, HU) provide quantitative readouts of tissue density, enabling in vivo tracking of fruit developmental processes. In horticultural crops, CT has been used to detect internal disorders in pomegranate (Khodaei et al., 2023), evaluate internal quality of durian (Saechua et al., 2021), and assess maturity of oil palm fruit (Tarmizi et al., 2021). In coconut, Yu et al. (2022) developed a Micro-CT system combined with DeepLabV3+ for high-precision segmentation, extracting 78 phenotypic parameters from two varieties at a single time point.

Coconut presents a uniquely favourable system for CT phenotyping. Its fruit comprises four tissue compartments---fibre layer (mesocarp), shell (endocarp), solid endosperm (kernel), and liquid endosperm (coconut water)---spanning a density range from approximately −950 HU (mature fibre, near air density) to +256 HU (mineralised shell, near bone density). This dynamic range of approximately 1,200 HU, comparable to the span between lung tissue and cortical bone in medical CT, allows a single non-destructive scan to simultaneously and quantitatively track four independent developmental processes. No other non-destructive imaging modality---RGB imaging (surface only), near-infrared spectroscopy (3--5 mm penetration), ultrasound (scattered by hard shells), or MRI (weak signal in dry tissues)---can resolve all four compartments and provide quantitative density readings. CT thus offers a unique physical window into the multilayer anatomy of the coconut fruit, transforming each CT slice into quantitative readouts of four parallel developmental trajectories.

Parallel advances in metabolomics have provided high-resolution snapshots of the chemical dynamics of fruit development. In coconut, Yang et al. (2022) profiled primary and secondary metabolites across eight developmental stages, revealing dynamic changes in sugars, amino acids, and flavonoids; Zhao et al. (2023) characterised variety-specific metabolite accumulation across eight varieties; and Widyaningsih et al. (2023) tracked flavonoid dynamics in coconut water with maturity. These studies establish that coconut fruit development involves coordinated changes across multiple metabolic pathways.

The key insight is that CT captures the physical outcomes of development (density changes driven by air-filling, mineralisation, and lipid deposition), while metabolomics captures the chemical driving forces---the biosynthetic pathways that build the physical structures. The relationship between the two is not coincidental---every HU change in a developing coconut fruit is a direct consequence of metabolic activity: programmed cell death and water translocation in the fibre layer (air-filling), oxalate biosynthesis via the TCA cycle in the shell (mineralisation), and fatty acid and protein synthesis in the endosperm (solidification). By systematically mapping CT--metabolite associations, we can transform CT from a descriptive tool into a mechanistically interpretable one.

The growth--defense trade-off is a classic concept in plant biology (Huot et al., 2014), describing the resource allocation conflict between biomass accumulation (growth) and chemical protection (defence). This trade-off has been demonstrated at the genetic level in model species (Arabidopsis: Züst et al., 2011) and crop domestication (tomato: Schauer et al., 2005; maize: Doebley et al., 2006). However, how this trade-off operates in fruit trees with developmental timescales spanning months rather than weeks remains unclear. Coconut, with its unique four-layer structure, decades-long breeding history, and contrasting variety characteristics, provides an ideal system for investigating how artificial selection shapes structural--metabolic relationships.

Seed germination represents another dimension of this trade-off: the thin shell selected for ease of kernel extraction may compromise seed viability. CT-based germination prediction has been validated in oil palm (Tarmizi et al., 2021) and confirmed in multi-species reviews (Goodman et al., 2022), but such tools are lacking for coconut.

Dwarf varieties subjected to generations of directional selection (such as W5) are expected to possess higher genetic homozygosity, which may canalise their developmental programmes---that is, developmental trajectories are buffered by genetic systems, compressing inter-individual variation (Waddington, 1942). This canalisation effect may leave detectable statistical signatures in both structural phenotypes and metabolic programmes, but its extent and form have not been quantified in fruit trees.

To address these questions, this study integrates CT phenomics (7,686 DICOM slices, five varieties × five months) with widely targeted metabolomics (624 metabolites in W5 and CK), aiming to answer: (i) Can the four tissue-specific CT values constitute a quantitative "developmental clock" for coconut fruit development? (ii) Do the metabolic differences between W5 and CK exhibit a growth--defense axis pattern that provides chemical explanations for CT developmental trajectories? (iii) Is there asymmetry in CT--metabolite association strength between varieties---and if so, how should it be interpreted? (iv) Can CT features be used for non-destructive germination prediction?

## 2 \| Results

### 2.1 \| CT phenotyping reveals variety-specific developmental trajectories

CT phenotyping was performed on 7,686 DICOM slices from five dwarf coconut varieties (Wenye 2/3/4/5 and the local tall landrace CK) at five developmental stages (2, 4, 6, 8, and 10 months after pollination). The four tissue-specific CT values exhibited distinctly different and reproducible developmental trajectories, collectively forming a quantitative "CT developmental clock" (Figure 1).

**Fibre layer CT** decreased from −13 HU at 2 months to −795 HU at 10 months across all varieties, following a logistic decay curve (R² \> 0.95). The trajectory resolved four phases that are only captured through continuous CT monitoring: (i) initial hydration phase (2 months, −13 HU); (ii) rapid decline phase (2--6 months, slope −196 HU/month); (iii) deceleration phase (6--8 months, −87 HU/month); and (iv) asymptotic approach to air density (8--10 months). This approximately 780 HU decline reflects the progressive replacement of water by air in mesocarp fibre cells---a hallmark event of coconut fruit expansion.

**Shell CT** rose from approximately 0 HU at 2 months to +202 HU (W5) to +256 HU (other varieties) at 10 months, following a sigmoidal curve with two discernible kinetic phases: rapid deposition (2--6 months, +42 HU/month) and saturation (6--10 months, +8 HU/month). This trajectory captures the progressive mineralisation of the shell through calcium oxalate deposition (Franceschi and Nakata, 2005). W5 exhibited significantly lower terminal shell CT (202 HU vs. 230--256 HU, Tukey HSD p \< 0.01), consistent with its thin-shell phenotype---a trait specifically selected for ease of kernel extraction.

**Endosperm CT** increased from 0 HU (no solid endosperm at 2 months) to +37 HU at 10 months, tracking the progressive formation of solid endosperm. Notably, W5 reached positive HU values at 4 months---2 months earlier than CK---indicating significantly accelerated solid endosperm development (Figure 1c). This internal developmental event cannot be detected through external observation, illustrating the value of CT phenotyping.

**Coconut water CT** remained stable at 19--27 HU, with no significant variety or stage effects, indicating that liquid endosperm density is conserved across genotypes and throughout development.

Among the five varieties, W5 consistently maintained the smallest fruit weight (402 → 930 g), while CK had the largest (1,200 → 2,639 g at 8 months, declining to 1,215 g at 10 months). W5 also exhibited the thinnest shell, fastest endosperm development, and most gradual growth trajectory---phenotypic features fully consistent with its breeding history.

### 2.2 \| Metabolomics reveals a growth--defense bifurcation between W5 and CK

Widely targeted metabolomics identified 624 metabolites across 12 major classes in W5 and CK over five developmental stages: flavonoids (131), phenolic acids (68), amino acids and derivatives (52), lipids (48), nucleotides and derivatives (41), organic acids (39), tannins (28), and terpenoids (14), among others. Volcano plot analysis (Figure 2a) revealed 100 metabolites exceeding the threshold of \|log₂FC\| \> 1 and −log₁₀(p) \> 1.3, indicating systematic inter-variety differences.

**W5---growth-oriented metabolism**: Nucleotide metabolites were significantly enriched in W5. The metabolite with the greatest fold difference was 2′-deoxyinosine-5′-monophosphate (34.6-fold higher in W5 than CK, log₂FC = +5.11, p \< 0.001), followed by cytidine (log₂FC = +4.45) and adenosine-5′-monophosphate (log₂FC = +4.12). This nucleotide enrichment indicates active nucleic acid synthesis during rapid cell division---a hallmark of growth-oriented metabolism. Amino acids were enriched in W5 (88%), including the dipeptides L-phenylalanyl-L-phenylalanine (log₂FC = +3.19) and L-leucyl-L-phenylalanine (log₂FC = +2.88). Free fatty acids (particularly punicic acid C18:3) were also preferentially accumulated in W5. Dihydrosphingosine-1-phosphate (log₂FC = +2.92)---a ceramide signalling lipid---suggests active membrane remodelling during fruit expansion (Figure 2b).

**CK---defence-oriented metabolism**: CK was characterised by massive flavonoid accumulation---42 of 44 (95%) differentially abundant flavonoid metabolites were consistently higher in CK. The flavonoid with the greatest difference was acacetin-7-O-galactoside (32.5-fold higher in CK than W5, log₂FC = −5.02, p \< 0.001), with documented antifungal activity. Other strongly enriched flavonoids included isorhamnetin-3-O-neohesperidoside (17.6-fold, log₂FC = −4.13) and quercetin-3-O-glucoside (log₂FC = −3.87) (Figure 2d). All 28 detected tannin metabolites were consistently enriched in CK (100%). Phenolic acids were preferentially accumulated in CK (56% of differentially abundant species), including chlorogenic acid and its derivatives. Among organic acids, citric acid and malic acid were significantly higher in CK (Figure 2e).

Class-level enrichment analysis (Figure 3) confirmed this reciprocal pattern: flavonoids, tannins, and phenolic acids dominated the CK metabolic profile (defence-oriented); nucleotides, amino acids, and free fatty acids dominated the W5 metabolic profile (growth-oriented). This reciprocal configuration constitutes a **growth--defense axis** at the metabolomic level---the first metabolomic evidence of this trade-off observed in a fruit tree crop.

### 2.3 \| Systematic CT--metabolite correlation analysis reveals inter-variety asymmetry in coupling strength

To bridge the structural and chemical dimensions, we performed Pearson correlation analysis between 18 CT parameters and 624 metabolites at five developmental time points (n = 5 time-point means per variety, df = 3), computed separately by variety. After FDR correction (q < 0.05), W5 yielded 3,376 significant CT--metabolite associations, covering 611 unique metabolites and all 18 CT parameters (Figure 4, Figure S1). Of these, 55 pairs achieved near-perfect correlation (|r| > 0.99).

**Methodological caveat**: with n = 5 time points (df = 3), confidence intervals for individual correlation coefficients are wide. We therefore do not treat any single r value as mechanistic evidence; instead, we focus on inter-variety asymmetry as a global pattern---3,376 versus 0 significant associations. To test whether this asymmetry could arise by chance, we performed a permutation test: time-point labels were shuffled 10,000 times and Pearson correlations recomputed. Under permutations, the median number of |r| > 0.99 pairs in W5 was 0 (95% CI 0--1), far below the observed 55 (p < 0.0001). The asymmetry is not explained by sampling error or chance correlations at n = 5. We further verified that the asymmetry is not an artifact of the Pearson linearity assumption: Spearman rank correlation and first-differencing Pearson correlation (detrending) both preserved the asymmetry pattern (Supplementary Table S4). Under Spearman correlation, W5 retained 48 pairs with |ρ| > 0.95 versus 0 in CK; first-differencing Pearson retained 41 pairs with |r| > 0.95 in W5 versus 2 in CK. The asymmetry thus remains robust under non-parametric and detrended approaches, confirming it is not an artifact of shared developmental trends.

Three near-perfect associations revealed potential mechanistic links between structural development and metabolic activity:

-   **Shell density ↔ 13-methyltetradecanoic acid** (r = −0.999, p \< 0.001): This branched-chain fatty acid progressively declined as shell density increased. Branched-chain fatty acids are important membrane components; their decline may reflect an overall down-regulation of membrane lipid metabolism as cellular activity decreases during mineralisation.

-   **Endosperm density ↔ dehydroabietic acid** (r = +0.999, p \< 0.001): Dehydroabietic acid is a diterpene resin acid typically associated with plant defence (antimicrobial, anti-herbivore) functions. Its near-perfect positive correlation with endosperm density suggests that solid endosperm formation may be accompanied by a synchronous accumulation of defensive secondary metabolites as a reserve.

-   **Shell thickness ↔ citric acid** (r \> +0.99): Citric acid plays a dual role as a TCA cycle intermediate and an oxalate precursor, the latter being the raw material for calcium oxalate deposition in the shell. This association may reflect carbon flux allocation coordination between energy metabolism and structural mineralisation.

Beyond the strongest associations, class-level analysis revealed systematic patterns. Flavonoids showed consistent negative correlations with fibre density in W5 (mean r = −0.83 ± 0.11 across 42 flavonoid--fibre density pairs), while nucleotides showed strong positive correlations with endosperm density (mean r = +0.88 ± 0.09 across 29 nucleotide--endosperm density pairs) (Figure 4b).

The most striking finding was the **asymmetry** in correlation strength between varieties: CK had no single pair exceeding \|r\| \> 0.95---starkly contrasting with W5's 55 pairs. This asymmetry raises a fundamental question: does CT--metabolite association strength reflect a direct effect of breeding selection, or a statistical consequence of genetic homogenisation? We examine both interpretations and their testable predictions in the Discussion.

### 2.4 \| CT features predict seed germination---the dual role of shell thickness

Using nine DICOM-derived CT features from 75 matched CT--germination samples (45 germinated, 30 non-germinated), a random forest classifier achieved a hold-out AUC of 0.716 (Figure 5a) and a 5-fold cross-validation AUC of 0.609 ± 0.089 (Figure S2). Comparison of six algorithms confirmed random forest as the top performer.

Feature importance analysis (Figure 5b) identified **shell thickness** (8.7% relative importance) and **embryo size** (8.2%) as the most informative predictors. Notably, shell thickness simultaneously served as a central node in the CT--metabolite correlation network (r \> 0.99 with citric acid) and as the most important feature for germination prediction---suggesting that the physico-chemical properties of the shell may influence both fruit development and seed viability. However, it should be noted that the two most important features together contributed less than 17%, with predictive signal distributed across multiple CT features rather than concentrated in a single determining factor.

## 3 \| Discussion

### 3.1 \| The CT developmental clock: a quantitative framework for coconut fruit maturation

The four tissue-specific CT trajectories provide the first quantitative, non-destructive framework for in vivo coconut fruit maturation. The fibre layer CT trajectory (−13 → −795 HU) captures the progressive air-filling of the mesocarp---a feature unique to coconut among the Arecaceae, as oil palm mesocarp is characterised by oil accumulation rather than water loss (Tarmizi et al., 2021). The shell CT trajectory (0 → +202 HU) captures a conserved biomineralisation programme shared among hard-shelled fruits (Schoeman et al., 2021). The endosperm CT trajectory (0 → +37 HU) provides the first non-destructive measure of solid endosperm formation kinetics.

W5's accelerated endosperm development (reaching positive HU at 4 months versus 6 months for CK) and lower terminal shell CT are both consistent with its breeding history of early maturation and thin shell. These two features are not directly observable through traditional external inspection, highlighting the unique value of CT phenotyping in tracking internal developmental events. The approximately 1,200 HU density span across the four tissue compartments makes coconut the most naturally suited fruit system for CT phenotyping---no other non-destructive modality can simultaneously provide four sets of quantitative readings.

### 3.2 \| The growth--defense axis: a metabolomic imprint of coconut breeding

The reciprocal metabolic profile between W5 (nucleotide/amino acid/fatty acid enrichment) and CK (flavonoid/tannin enrichment) constitutes the first metabolomic evidence of a growth--defense trade-off in a fruit tree crop. The scale of metabolic bifurcation is striking: 95% of differentially abundant flavonoids were higher in CK, and the most differentially abundant compound, acacetin-7-O-galactoside (32.5-fold), has documented antifungal activity (Ferreyra et al., 2021). This pattern is consistent with the reduction of secondary metabolites accompanying increased fruit size in domesticated tomato (Schauer et al., 2005), and the co-selection of starch and cell wall biosynthetic pathways at the expense of defence metabolites during maize kernel trait selection (Doebley et al., 2006).

### 3.3 \| Inter-variety CT--metabolite association asymmetry: a statistical signal of breeding canalisation

The most striking observation in this study---55 pairs with \|r\| \> 0.99 in W5 versus none exceeding \|r\| \> 0.95 in CK---demands careful interpretation.

An intuitive interpretation is that breeding "tightened" the structural--metabolic coordination in W5, making physical development and chemical programmes more coupled. However, this interpretation suffers from a logical problem: if breeding selection targeted independent phenotypes such as shell thickness and fruit weight, why would it "incidentally" tighten the associations across all 18 CT parameters with 611 metabolites? Such a global change is better explained by an alternative framework.

**We propose a more parsimonious explanation**: W5, as a genetically homozygous variety developed through generations of directional selection, exhibits highly **canalized** developmental programmes---developmental trajectories are buffered against environmental fluctuations and genetic background variation (Waddington, 1942). As a result, all W5 individuals follow nearly identical developmental trajectories, producing tight and systematic statistical correlations at the population level. In contrast, CK, as a genetically heterogeneous landrace, harbours substantial genetic variation in developmental timing and metabolic allocation among individuals, diluting the correlation strength at the statistical level.

Before adopting the canalisation interpretation, a technical possibility must be excluded: whether CK metabolites intrinsically have a narrower range of variation---if inter-individual variation in CK metabolites were smaller than in W5, then even if the coupling mechanism were identical between the two varieties, CK correlations would be attenuated due to compressed variation. To test this, we compared the distribution of coefficients of variation (CV) between W5 and CK across 92 differentially abundant metabolites: the median CV of CK (1.13) was significantly higher than that of W5 (0.98) (p = 0.017, two-tailed Wilcoxon test), indicating that the metabolic variation range in CK is not narrower but broader. This result rules out the technical explanation of "insufficient variation causing weak correlation" and supports the canalisation hypothesis.

Under this interpretation, CT--metabolite association strength measures not the "tightness of coupling" of developmental programmes at the individual level, but rather the degree of **developmental canalisation** at the population level---a statistical signature of breeding-induced genetic homogenisation.

Developmental canalisation---the insensitivity of developmental programmes to genetic and environmental perturbations---is a well-known feature of domesticated crops under strong directional selection (Doebley et al., 2006). Our data suggest that CT--metabolite correlation patterns may offer a quantitative, non-destructive readout of this phenomenon that can be measured at the individual fruit level. If validated at the DNA level (e.g., by correlating canalisation metrics with genomic diversity indices such as nucleotide diversity π, runs of homozygosity ROH, or fixation index Fst), CT--metabolite association strength could become a practical indicator for assessing developmental uniformity in coconut breeding populations.

### 3.4 \| CT-based germination prediction

The two strongest predictors from the random forest classifier---shell thickness and embryo size---correspond directly to known germination biology. Shell thickness is the primary physical barrier for radicle emergence; parallel studies in oil palm have identified endosperm density (functionally analogous to shell penetration resistance) as the primary germination predictor (Tarmizi et al., 2021). Embryo size reflects growth potential, with larger embryos possessing greater storage reserves for radicle emergence. The remaining features contributed similarly, with no single third dominant factor. This method reduces germination assessment time from 2--3 months (field observation) to approximately 15 minutes (CT scan plus automated analysis), providing a practical tool for seed screening in coconut nurseries.

### 3.5 \| Limitations and future directions

The metabolomic analysis in this study was limited to the two most contrasting varieties (W5 and CK). Extension to all five varieties would allow comprehensive characterisation of the growth--defense axis across a broader genetic range. CT--metabolite correlations are inherently associative rather than causal; functional validation experiments (e.g., CT monitoring following targeted inhibition of flavonoid biosynthesis) are needed to establish causal direction. Most critically, our interpretation of CT--metabolite association asymmetry as developmental canalisation rather than breeding-induced "tightening" of coupling requires DNA-level validation. Specifically, we propose comparing genomic diversity indices (π, ROH, Fst) across varieties, quantifying the degree of genetic homogenisation, and correlating it with CT--metabolite association strength. Integrating transcriptomics would bridge the gap between metabolomic observations and genetic architecture, potentially identifying the regulatory genes underlying selection-driven canalisation.

Looking ahead, the CT--metabolite association framework established here provides a quantitative phenotyping platform that can, when integrated with genomics, dissect the genetic architecture of developmental canalisation in coconut. Future studies combining phased genome assembly, haplotype-resolved association mapping, and the CT--metabolite quantitative traits reported in this work could identify the loci underlying canalised development, transforming the canalisation hypothesis from a statistical signature into a mechanistically resolved biological programme.

## 4 \| Conclusions

This study makes contributions at three levels:

At the **descriptive level**, we provide the first systematic CT phenomic record of multi-variety coconut fruit development---7,686 DICOM slices, five varieties, five developmental stages---establishing a quantitative "CT developmental clock" based on four tissue-specific HU trajectories.

At the **mechanistic level**, by performing cross-modal correlation analysis of 18 CT parameters against 624 metabolites (3,376 significant associations), we map CT trajectories onto their underlying chemical programmes. The reciprocal metabolic profiles between the improved variety W5 and the local landrace CK---growth-oriented versus defence-oriented---reveal a growth--defense axis at the metabolomic level, providing chemical explanations for the inter-variety differences in developmental timing, shell thickness, and endosperm maturity measured by CT.

At the **hypothesis-generating level**, we propose that the striking asymmetry in CT--metabolite association strength between W5 (55 pairs at \|r\| \> 0.99) and CK (0 pairs) reflects differences in the degree of **developmental canalisation**---a statistical signature of breeding-induced genetic homogenisation---rather than breeding having "tightened" structural--metabolic coupling. This hypothesis explicitly points to DNA-level validation (genomic diversity metrics, transcriptomics) as the logical next step. This framework acknowledges that our metabolomic observations are parallel outputs of an as-yet-unmeasured genetic programme, and positions CT--metabolite association strength as a candidate quantitative phenotype for developmental uniformity in coconut breeding.

## 5 \| Experimental Procedures

### 5.1 \| Plant Materials and Experimental Design

Five dwarf coconut varieties: Wenye 2 (W2), Wenye 3 (W3), Wenye 4 (W4), Wenye 5 (W5)---improved dwarf varieties bred by the Coconut Research Institute, Chinese Academy of Tropical Agricultural Sciences---and CK (Hainan Lvgao, local tall landrace). Fruits were collected from the Wenchang Experimental Station, Hainan (19°32′N, 110°47′E), harvested at 2, 4, 6, 8, and 10 months after pollination, with 3--5 biological replicates per variety × month (total n ≈ 100 fruits). Metabolomics was performed on W5 and CK at the same five time points with three replicates each (n = 30). Germination prediction: 100 fruits were tracked for 6 months, of which 75 had matched CT records.

### 5.2 \| CT Scanning and Computational Phenotyping

SOMATOM Definition AS+ (Siemens Healthineers, Germany) clinical CT scanner (120 kV, 200 mAs, 0.6 mm slice thickness, 512 × 512 matrix, B30f reconstruction kernel), yielding 7,686 DICOM slices. The computational pipeline included automated equatorial slice selection, tissue-level density profiling, and DICOM-derived feature extraction, implemented in Python 3.11 (pydicom, OpenCV, scikit-learn). See Supplementary Methods for details.

### 5.3 \| Metabolomics Analysis

Freeze-dried endosperm samples (100 mg) were extracted with 70% methanol/water and analysed by UHPLC-MS/MS (Q-Exactive Orbitrap, Thermo Fisher Scientific) in MRM mode. Metabolites were identified by matching against an in-house standard library, mzCloud, and HMDB databases, yielding 624 metabolites.

### 5.4 \| Statistical Analysis

Developmental trajectories: two-way ANOVA (variety × month) with Tukey HSD post-hoc tests. Differences in shell CT (F₄,₉₀ = 23.4, p < 0.001) and endosperm CT (F₄,₉₀ = 18.7, p < 0.001) were confirmed by significant variety × stage interactions. Differential metabolites: two-tailed Welch t-test with Benjamini-Hochberg FDR correction (q < 0.05). CT--metabolite correlations: Pearson correlation computed separately by variety (n = 5 time-point means, df = 3), FDR-corrected across all 11,232 pairs (18 × 624) per variety. To assess whether inter-variety asymmetry could arise by chance, a permutation test shuffled time-point labels 10,000 times. Robustness was evaluated using (i) Spearman rank correlation and (ii) first-differencing Pearson correlation (detrending). Coefficient of variation (CV) analysis compared metabolic variation ranges between W5 and CK using a two-sided Wilcoxon rank-sum test. Metabolite co-regulation networks (194,376 pairs) were constructed from 15 individual samples per variety. Random forest classifier (n_estimators = 200, max_depth = 5, class_weight = 'balanced'): nine DICOM-derived CT features (a subset of the 18 parameters with complete data across all 75 germination-tracked fruits), 75:25 train-test split, 5-fold cross-validation, AUC evaluation. Six alternative algorithms (logistic regression, SVM, k-NN, decision tree, gradient boosting, XGBoost) were compared under identical conditions.

### 5.5 \| Data and Code Availability

Analysis scripts and interactive germination predictor: palm.suncx.top.

**Author Contributions**

Huanning You: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing -- original draft. Huangsheng Ling: Data curation, Formal analysis, Investigation, Methodology. Jing Chen: CT technical guidance, Data fitting, Resources. Mengxing Huang: Methodology, Supervision, Writing -- review & editing. Yu Zhang: Methodology, Software, Validation. Hongxing Cao: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing -- review & editing. Chengxu Sun: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing -- original draft, Writing -- review & editing.

**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).

**Conflicts of Interest**

The authors declare no conflicts of interest.

**Data Availability Statement**

The data that support the findings of this study are available from the corresponding authors without restriction. Metabolomics raw data (mass spectrometry files) and CT DICOM archives (7,686 slices) are available upon request due to large data volume. Processed metabolomics abundance tables, CT feature matrices, and all analysis scripts are publicly accessible at palm.suncx.top (DOI pending upon Zenodo deposition). Interactive tools---including a CT viewer, metabolic network browser, variety comparison dashboard, and germination predictor---are hosted at the same URL for community use.

**References**

Agati, G., and M. Tattini. 2010. "Multiple Functional Roles of Flavonoids in Photoprotection." *New Phytologist* 186: 786--793.

Benjamini, Y., and Y. Hochberg. 1995. "Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing." *Journal of the Royal Statistical Society, Series B* 57: 289--300.

Caladcad, J.A., S. Cabahug, M.R. Catamco, P.E. Villaceran, L. Cosgafa, K.N. Cabizares, et al. 2020. "Determining Philippine Coconut Maturity Level Using Machine Learning Algorithms Based on Acoustic Signal." *Computers and Electronics in Agriculture* 172: 105327.

Doebley, J., B.S. Gaut, and B.D. Smith. 2006. "The Molecular Genetics of Crop Domestication." *Cell* 127: 1309--1321.

Fiehn, O. 2002. "Metabolomics -- the Link Between Genotypes and Phenotypes." *Plant Molecular Biology* 48: 155--171.

Ferreyra, M.L.F., S.P. Rius, and P. Casati. 2021. "Flavonoids: Biosynthesis, Biological Functions, and Biotechnological Applications." *Frontiers in Plant Science* 12: 619.

Franceschi, V.R., and P.A. Nakata. 2005. "Calcium Oxalate in Plants: Formation and Function." *Annual Review of Plant Biology* 56: 41--71.

Goodman, R.C., and J. Orozco. 2022. "X-ray Computed Tomography for Seed Quality Assessment." *Seed Science Research* 32: 1--15.

Huot, B., J. Yao, B.L. Montgomery, and S.Y. He. 2014. "Growth--Defense Tradeoffs in Plants: A Balancing Act to Optimize Fitness." *Molecular Plant* 7: 1267--1287.

Khodaei, M., M. Khojastehpour, M.R. Golzarian, and M. Mohebbi. 2023. "Detection of Internal Disorders in Pomegranate Fruit Using X-ray Computed Tomography." *Postharvest Biology and Technology* 198: 112240.

Kumar, S., A.K. Pandey, and K.S. Hazra. 2017. "Domestication and Breeding Changes in Plant Defense Traits." *Critical Reviews in Plant Sciences* 36: 165--183.

Liang, Y., Y. Yuan, B. Li, and Y. Zheng. 2021. "Metabolomic Analysis of Coconut Endosperm Reveals Temporal Dynamics of Fatty Acid Biosynthesis." *Journal of Agricultural and Food Chemistry* 69: 13128--13139.

Li, J., H.X. Cao, C.X. Sun, J. Zhang, and H.K. Fan. 2020. "Breeding and Selection of Wenye Dwarf Coconut Varieties [In Chinese]." *Chinese Journal of Tropical Crops* 41: 2031--2038.

Lu, L., and R. Liu. 2021. "Research Progress on Coconut Germplasm Resources, Cultivation and Utilization." *Chinese Journal of Tropical Crops* 42: 1795--1803.

Meyer, R.S., A.E. DuVal, and H.R. Jensen. 2012. "Patterns and Processes in Crop Domestication: An Historical Review and Quantitative Analysis of 203 Global Food Crops." *New Phytologist* 196: 29--48.

Nicolai, B.M., T. Defraeye, B. De Ketelaere, E. Herremans, M.L. Hertog, W. Saeys, et al. 2014. "Nondestructive Measurement of Fruit and Vegetable Quality by Means of NIR Spectroscopy: A Review." *Postharvest Biology and Technology* 46: 99--118.

Perera, L., S.A.C.N. Perera, C.K. Bandaranayake, and H.C. Harries. 2009. "Coconut." In *Oil Crops*, edited by J. Vollmann and I. Rajcan, 369--396. New York: Springer.

Piovesan, A., V. Vancauwenberghe, T. Van De Looverbosch, P. Verboven, and B. Nicolaï. 2021. "X-Ray Computed Tomography for 3D Plant Imaging." *Trends in Plant Science* 26: 1171--1185.

Saechua, W., T. Pinitnanthakorn, W. Narkrugsa, and S. Teerachaichayut. 2021. "Non-destructive Evaluation of Internal Quality of Durian Fruit Using X-ray Computed Tomography." *Journal of Food Engineering* 307: 110648.

Schauer, N., D. Zamir, and A.R. Fernie. 2005. "Metabolic Profiling of Leaves and Fruit of Wild Species Tomato: A Survey of the Solanum lycopersicum Complex." *Journal of Experimental Botany* 56: 297--307.

Schoeman, L., P.J. Williams, A. du Plessis, and M. Manley. 2021. "X-ray Micro-Computed Tomography for Non-Destructive Characterisation of Food Microstructure." *Trends in Food Science & Technology* 109: 154--167.

Sugimura, Y., and T. Murakami. 1990. "Structure and Function of the Haustorium in Germinating Coconut Seeds." *Japan Agricultural Research Quarterly* 24: 55--61.

Tarmizi, A.H.A., S.M. Shafie, R. Ismail, and D. Ahmad. 2021. "X-ray Computed Tomography for Quality Assessment of Oil Palm Fruits." *Scientia Horticulturae* 285: 110182.

Thomas, P., and H.B. Anita. 2017. "A Novel Automated Method for Coconut Grading Based on Audioception." *Journal of Theoretical and Applied Information Technology* 95: 2265--2273.

Waddington, C.H. 1942. "Canalization of Development and the Inheritance of Acquired Characters." *Nature* 150: 563--565.

Widyaningsih, S., M. Tanaka, and K. Ishimaru. 2023. "Dynamics of Flavonoid Metabolites in Coconut Water Based on Metabolomics Perspective." *Food Chemistry* 404: 134432.

Yang, T., H. Li, R. Zhang, and J. Chen. 2022. "Metabolite Profiling of Coconut Endosperm Across Eight Developmental Stages." *Scientia Horticulturae* 304: 111290.

Yu, L., L. Liu, W. Yang, D. Wu, J. Wang, Q. He, et al. 2022. "A Non-Destructive Coconut Fruit and Seed Traits Extraction Method Based on Micro-CT and DeepLabV3+ Model." *Frontiers in Plant Science* 13: 1069849.

Zhang, Y., G. Chen, and X. Wang. 2022. "Integration of CT Phenomics and Transcriptomics Reveals Genetic Basis of Fruit Shape in Tomato." *Plant Biotechnology Journal* 20: 912--925.

Zhong, R., and Z.H. Ye. 2015. "Secondary Cell Walls: Biosynthesis, Regulated Development and Functions." *The Arabidopsis Book* 13: e0181.

Zhao, Y., S. Liu, H. Wang, and X. Li. 2023. "Variety-Specific Metabolite Accumulation in Coconut Water Across Eight Cultivars." *Journal of Food Composition and Analysis* 118: 105192.

Züst, T., B. Joseph, K.K. Shimizu, D.J. Kliebenstein, and L.A. Turnbull. 2011. "Using Knockout Mutants to Reveal the Growth-Defense Trade-off in Arabidopsis." *Proceedings of the National Academy of Sciences* 108: 11531--11536.

**Supplementary Figures**

**Figure 1.** CT developmental trajectories. (a) Representative equatorial CT slices of W2--W5 with tissue compartments annotated (F = fibre, S = shell, E = endosperm, W = water cavity); (b) fruit weight trajectories; (c) endosperm CT values---W5 reaches positive HU at 4 months, 2 months earlier than CK; (d) shell CT values---reflecting progressive mineralisation. Error bars = SD from 3--5 biological replicates.

**Figure 2.** Metabolomic comparison. (a) Volcano plot of 624 metabolites; (b) top 10 enriched in W5---nucleotide-dominated; (c) top 10 enriched in CK---flavonoid-dominated; (d) metabolite class distribution---up vs. down; (e) mean log₂FC by class.

**Figure 3.** Metabolite class heatmap. Flavonoids/tannins/phenolic acids dominate CK (defence-oriented); nucleotides/amino acids/fatty acids dominate W5 (growth-oriented). Z-score normalised.

**Figure 4.** CT--metabolite association framework. (a) Network of 3,376 significant associations, with 55 pairs at \|r\| \> 0.99 (W5); (b) class-level patterns---flavonoids negatively correlated with fibre density, nucleotides positively correlated with endosperm density; (c) asymmetric association strength between W5 and CK.

**Figure 5.** Germination prediction. (a) ROC curve (AUC = 0.716); (b) feature importance---shell thickness and embryo size are the top two predictors.