# Source paper archive

- Source title: A super-pangenome of Capsicum reveals genomic basis of domestication, pungency, and metabolic diversity
- Author(s): Zhao, J., Xiong, C., Zhao, X., Zhao, P., Sun, H., Yu, L., Liu, Y., Jarret, R., Zhu, F., Dai, X., Yang, S., Liu, Z., Yu, H., Ou, L., Yang, B., Xu, H., Li, Y., Wang, Z., Hu, B., Weng, R., Zou, X., Fei, Z., Liu, F.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.17.767367v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.17.767367v1.full.pdf
- PDF SHA-256: f6307a934d7ad5517fb5aa2a5cef7366bde5825d2f2d2028b34977dfee103090
- Converter: pymupdf4llm 0.0.27
- Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version.

---

&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**A super-pangenome of** _**Capsicum**_ **reveals genomic basis of domestication, pungency, and**

**metabolic diversity**


Jiantao Zhao &#91;1,2,6,#&#93;, Cheng Xiong &#91;1,#&#93;, Xuebo Zhao &#91;2,#&#93;, Pei Zhao &#91;1,#&#93;, Honghe Sun &#91;2,#&#93;, Lujia Yu &#91;1&#93;, Yi Liu &#91;1&#93;,
Robert Jarret &#91;3&#93;, Fan Zhu &#91;1&#93;, Xiongze Dai &#91;1&#93;, Sha Yang &#91;1&#93;, Zhoubin Liu &#91;1&#93;, Huiyang Yu &#91;1&#93;, Lijun Ou &#91;1&#93;, Bozhi
Yang &#91;1&#93;, Hao Xu &#91;1&#93;, Yanlong Li &#91;1&#93;, Zhongyi Wang &#91;1&#93;, Bowen Hu &#91;1&#93;, Rui Weng &#91;4&#93; *, Xuexiao Zou &#91;1,5*&#93;,
Zhangjun Fei &#91;2*&#93;, Feng Liu &#91;1,5*&#93;

1 Engineering Research Center for Germplasm Innovation and New Varieties Breeding of
Horticultural Crops, Key Laboratory for Vegetable Biology of Hunan Province, College of
Horticulture, Hunan Agricultural University, Changsha, China
2 Boyce Thompson Institute, Cornell University, Ithaca, NY, USA
3 U.S. Department of Agriculture-Agricultural Research Service, Plant Genetic Resources
Conservation Unit, Griffin, GA, USA.
4 State Key Laboratory for Quality and Safety of Agro-Products, Key Laboratory of Agro-food
Safety and Quality of Ministry of Agriculture and Rural Affairs, Institute of Quality Standard and
Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences, Beijing, China
5 Yuelushan Laboratory, Changsha, China
6 Present address: Key Laboratory of Genome Research and Genetic Improvement of Xinjiang
Characteristic Fruits and Vegetables, Institute of Fruits and Vegetables, Xinjiang Uygur
Autonomous Region Academy of Agricultural Sciences, Urumqi, China
# These authors contributed equally

**Abstract**

The _Capsicum_ genus encompasses remarkable genetic and phenotypic diversity underlying its
agricultural, economic, and cultural importance. Here, we present a _Capsicum_ super-pangenome
constructed from 48 chromosome-scale genomes, including 38 newly assembled, spanning all five
domesticated taxa and their wild relatives. We identified 72,687 pangene families and abundant
structural variants (SVs), revealing lineage-specific gene-content dynamics and patterns consistent
with strong purifying selection against large SVs. A pan-NLRome revealed extensive presenceabsence variation in the disease-resistance gene repertoire, with domesticated accessions harboring
fewer NLRs than their wild progenitors. The graph-based super-pangenome enabled extensive
genome-wide association mapping and robust genomic prediction across 130 agronomic and
metabolic traits, leading to the identification and functional validation of _RAP2_ as a regulator of
carotenoid accumulation and the discovery of a structurally complex terpene synthase locus
associated with sesquiterpene ylangene. Pangenome-wide dissection of capsaicinoid biosynthesis
further revealed evolutionary remodeling of the pathway through tandem gene duplication,
interchromosomal rearrangement, and regulatory specialization. Collectively, our findings
illuminate the genomic basis of _Capsicum_ domestication and metabolic diversity and establish a
comprehensive resource for genomics-assisted improvement of disease resistance, fruit quality,

and flavor.


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Introduction**

Pepper ( _Capsicum_ spp.) belongs to the genus _Capsicum_, which comprises approximately 35
species, including five domesticated taxa: _C. annuum_ var. _annuum_, _C. chinense_, _C. frutescens_, _C._
_baccatum_ var. _pendulum_, and _C. pubescens_ . Originating in the Americas (Kim et al., 2014),
peppers are now among the most widely cultivated crops worldwide and have profoundly shaped
diets and culinary traditions across diverse cultures (Katz, 2019). They also hold substantial
economic and cultural importance, with diverse applications in cuisine, medicine, and industry
(Hernández-Pérez et al., 2020). Genomic studies have begun to reveal how genetic diversity and
gene flow have shaped pepper domestication and population differentiation (Liu et al., 2023).
However, the genetic basis of key agronomic traits, including disease resistance, fruit morphology,
nutritional quality, and volatile composition, remains largely unexplored, particularly in
underrepresented domesticated taxa such as _C. pubescens_ and _C. baccatum_ . Addressing these
knowledge gaps requires a comprehensive framework capable of capturing the full spectrum of
genomic variation across _Capsicum_ species and linking this variation to agronomic and
metabolomic diversity.

Although more than 20 pepper genome assemblies have been published, many were
generated using short-read or earlier long-read sequencing technologies with lower accuracy (Kim
et al., 2014; Qin et al., 2014; Kim et al., 2017; Hulse-Kemp et al., 2018; Lee et al., 2022; Liao et
al., 2022). High-fidelity (HiFi) long-read sequencing has substantially improved genome assembly
accuracy, continuity, and completeness, enabling high-quality chromosome-scale assemblies
across diverse pepper species (Shirasawa et al., 2023; Liu et al., 2023; Delorean et al., 2023; Chen
et al., 2024; Zhang et al., 2025). In parallel, recent advances in pangenomics provide a powerful
framework for characterizing genetic diversity within and across species. By integrating multiple
high-quality genome assemblies, pangenomes capture a more complete repertoire of genetic
variation than a single reference genome, thereby providing deeper insights into evolutionary
dynamics, population diversity, and functional variation (Wang et al., 2023; Schreiber et al., 2024).
Furthermore, graph-based pangenomes‒which represent multiple genome sequences within an
interconnected graph‒have been shown to substantially improve the detection and genotyping of
variants at the population scale, particularly large structural variants (SVs), compared to
conventional linear reference approaches. These advances have facilitated trait discovery and
genomics-assisted breeding in various crops (Zhou et al., 2022; Wang et al., 2023; Schreiber et al.,
2024; Sun et al., 2026; Zhao et al., 2026).
Here, we present a _Capsicum_ super-pangenome constructed from high-quality genome
assemblies of 48 accessions, including 38 newly generated genomes, spanning all five
domesticated taxa and their wild relatives. Integrating this genomic resource with population-scale
genomic variation and extensive phenotypic characterization of 130 agronomic and metabolic
traits‒including morphology and growth, carotenoids, capsaicinoids, and fruit volatiles‒provides
insights into the genetic and functional diversity of the _Capsicum_ genus. Collectively, our results
illuminate the genomic basis of pepper domestication and metabolic diversification and establish
valuable resources for genomics-assisted pepper improvement.


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Results**

**Genome assembly and annotation of** _**Capsicum**_ **accessions**
To comprehensively capture genomic diversity across the _Capsicum_ genus, we selected 38
representative accessions, primarily based on phylogenetic relationships among 500 core
accessions (Liu et al., 2023). These comprised 23 cultivated accessions (five _C. pubescens_, four _C._
_baccatum_ var. _pendulum_, five _C. frutescens_, four _C. chinense_, and five _C. annuum_ var. _annuum_ )
and 15 wild accessions (one _C. rhomboideum_, four _C. chacoense_, five _C. baccatum_ var. _baccatum_,
one _C. galapagoense_, and four _C. annuum_ var. _glabriusculum_ ). For each accession, we generated
an average of 93.2 Gb (~29×) of PacBio HiFi sequence data and 265.0 Gb (~83×) of highthroughput chromosome conformation capture (Hi-C) data. Using these datasets, we generated
chromosome-scale assemblies with an average contig N50 of 173.4 Mb ( **Supplementary Fig. 1**
and **Supplementary Table 1** ). Assembly sizes ranged from 3.03 to 3.97 Gb, except for _C._
_rhomboideum_, which had a substantially smaller genome (~1.7 Gb). Genome size was positively
correlated with transposable element (TE) content (Pearson’s _r_ = 0.63, _P_ = 1.1 × 10 &#91;-6&#93; ). Quality
assessments using Merqury (Rhie et al., 2020), BUSCO (Simão et al., 2015), and LAI (Ou et al.,
2018) confirmed the high accuracy and completeness of the assemblies ( **Supplementary Table 1** ).
We integrated the 38 newly generated assemblies with ten previously published high-quality
genomes (Lee et al., 2022; Liao et al., 2022; Shirasawa et al., 2023; Liu et al., 2023; Delorean et
al., 2023; Chen et al., 2024), yielding a dataset of 48 _Capsicum_ genomes spanning diverse genetic
groups: two _C. rhomboideum_, six _C. pubescens_, four _C. chacoense_, five _C. baccatum_ var. _baccatum_,
five _C. baccatum_ var. _pendulum_, one _C. galapagoense_, five _C. frutescens_, four _C. chinense_, four
_C. annuum_ var. _glabriusculum_, and 12 _C. annuum_ var. _annuum_ accessions ( **Fig. 1** and
**Supplementary Table 1** ). Genome-wide synteny was largely conserved across the 48 accessions,
despite several large-scale rearrangements among lineages ( **Fig. 1** ).

Gene prediction integrating _ab initio_ prediction, transcript evidence, and protein homology
identified an average of 37,005 protein-coding genes per genome. BUSCO analysis indicated that,
on average, 95.74% of conserved single-copy plant orthologs were identified as complete,
indicating high completeness of the gene predictions. For the ten previously published genomes,
gene models were updated using Liftoff (Shumate and Salzberg, 2021), achieving annotation
completeness comparable to that of the newly assembled genomes ( **Supplementary Table 1** ).

**Gene-based super-pangenome and pan-NLRome**
To characterize gene-content diversity across the _Capsicum_ genus, we constructed a gene-based
super-pangenome from the 48 genomes and identified 72,687 gene families (pangenes). The
pangene accumulation curve increased rapidly with the addition of the first several genomes and
approached saturation at approximately 35 accessions ( **Fig. 2a** ). Based on their occurrence
frequencies across the 48 genomes, pangenes were classified into four categories: core (10,427;
14.3%), soft-core (3,025; 4.2%), shell (55,952; 77.0%), and private (3,283; 4.5%) ( **Fig. 2b** ). The
predominance of shell pangenes indicates extensive gene presence-absence variation (PAV) across
_Capsicum_ . Functional enrichment analyses revealed that core pangenes were primarily involved


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Fig. 1** Phylogeny and genome synteny across the _Capsicum_ genus. **a,** Phylogenetic relationships among the

48 _Capsicum_ accessions included in this study. Fruit images are shown alongside the corresponding

accessions where available. **b,** Genome-wide synteny among the 48 _Capsicum_ genomes, generated using


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


GENESPACE (https://github.com/jtlovell/GENESPACE). Horizontal gray bars represent chromosomes,

and ribbons connecting adjacent genomes indicate syntenic relationships.

in essential cellular and metabolic processes, whereas soft-core, shell, and private pangenes were
enriched for functions related to environmental adaptation and stress responses ( **Supplementary**
**Fig. 2** ).

We reconstructed a time-calibrated phylogeny of _Capsicum_, together with tomato, potato,
and eggplant, using 5,020 single-copy orthologs. The phylogeny placed _C. rhomboideum_ as the
earliest-diverging lineage within _Capsicum_, followed by _C. pubescens_, with the remaining species
forming a more recently diversified clade. Divergence-time estimates indicated that _C._
_rhomboideum_ diverged from the remaining _Capsicum_ lineages approximately 13.56 million years
ago (Ma), followed by _C. pubescens_ at approximately 3.90 Ma, whereas the remaining _Capsicum_
lineages diversified approximately 1.24-2.42 Ma ( **Supplementary Fig. 3** ). Within this
phylogenetic framework, comparison of pangene distributions across major _Capsicum_ lineages
revealed that, although a substantial proportion of pangenes were shared among lineages, each
lineage also contained a distinct set of lineage-specific pangenes ( **Supplementary Fig. 4** ).
To investigate disease-resistance gene diversity, we constructed a pan-NLRome from the 48
_Capsicum_ genomes. In total, 6,533 NLR genes were identified and clustered into 1,115
orthogroups (pangenes). Most NLR pangenes were classified as shell or private ( **Fig. 2c** ), revealing
extensive variation in NLR repertoires among accessions. The pan-NLRome accumulation curve
approached saturation at approximately 40 accessions ( **Supplementary Fig. 5** ). NLR genes were
further classified into four subclasses: TIR-NLR (TNL), CC-NLR (CNL), CCR-NLR (RNL), and
NB-and-LRR-only (NL). Across all accessions, NLs represented the predominant subclass,
followed by CNLs, whereas TNLs and RNLs were comparatively rare ( **Fig. 2d** ). Comparison of
NLR repertories across major _Capsicum_ groups revealed broadly similar distributions of NLR
subclasses but substantial lineage-specific variation in NLR copy number ( **Fig. 2e** ). _C. pubescens_,
_C. baccatum_ var. _baccatum_, and _C. baccatum_ var. _pendulum_ contained relatively large NL
repertoires, whereas _C. rhomboideum_ possessed the fewest NL genes. CNLs were more abundant
in _C. pubescens_, _C. frutescens_, and _C. annuum_ var. _glabriusculum_ than in the other groups. TNLs
were most abundant in _C. rhomboideum_ and _C. pubescens_ but less abundant in several more
recently diverged lineages, including _C. baccatum_ var. _pendulum_, _C. galapagoense_, _C. frutescens_,
and _C. annuum_ var. _annuum_ . RNLs were consistently the least abundant subclass across all groups.
We further compared NLR repertoires along two independent domestication trajectories
represented in our dataset: _C. baccatum_ var. _baccatum_ versus domesticated _C. baccatum_ var.
_pendulum_, and _C. annuum_ var. _glabriusculum_ versus domesticated _C. annuum_ var. _annuum_ . In
both comparisons, domesticated accessions generally contained fewer NLR genes than their
corresponding wild relatives ( **Fig. 2e** ), suggesting that domestication was accompanied by a
reduction in NLR copy number.

Together, the gene-based super-pangenome and pan-NLRome provide a framework for
characterizing gene-content and immune-receptor diversity across _Capsicum_, facilitating
comparative analyses of genome evolution and providing a valuable resource for the genetic


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


improvement of agronomically important traits, particularly disease resistance.


**Fig. 2 Gene-based super-pangenome and pan-NLRome of the** _**Capsicum**_ **genus. a,** Pangenome

expansion and core-genome contraction with increasing numbers of genomes. For each given number of

genomes, 100 random combinations of accessions were sampled. Boxplots show the distributions across

random combinations, and dashed lines indicate the fitted accumulation and contraction curves for the

pangenome and core genome, respectively. **b,c,** Distribution of pangene ( **b** ) and NLR orthogroup ( **c** )

occurrence frequencies across the 48 genomes. Pangenes and NLR orthogroups were classified as core

(present in all 48 genomes), soft-core (present in 43–47 genomes), shell (present in 2–42 genomes), and

private (present in a single genome). Insets show the proportions of pangenes and NLR orthogroups in the

four categories. **d,** NLR copy numbers across the 48 _Capsicum_ accessions, with NLRs classified into NL,

CNL, TNL, and RNL subclasses. **e,** Comparison of NLR copy numbers among major _Capsicum_ groups.

For each boxplot, the lower and upper bounds indicate the first and third quartiles, respectively, the


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


centerline indicates the median and the whiskers extend to 1.5× interquartile range. CRH, _C. rhomboideum_ ;

PUB, _C. pubescens_ ; CHA, _C. chacoense_ ; BAB, _C. baccatum_ var. _baccatum_ ; BAP, _C. baccatum_ var.

_pendulum_ ; GAL, _C. galapagoense_ ; CHN, _C. chinense_ ; FRU, _C. frutescens_ ; GLA, _C. annuum_ var.

_glabriusculum_ ; ANN, _C. annuum_ var. _annuum_ .

**Graph-based super-pangenome construction and SV genotyping**
To comprehensively capture structural variation across _Capsicum_, we constructed a graph-based
super-pangenome using 46 high-quality genomes, excluding the two highly divergent _C._
_rhomboideum_ accessions. The graph contained a large set of SVs (≥20 bp), including 2,012,901
insertions and 1,498,594 deletions ( **Supplementary Table 2** ). SV abundance varied substantially
among accessions, ranging from 34,033 to 273,156 variants per accession ( **Supplementary Table**
**3** ), highlighting extensive structural diversity across _Capsicum_ . Functional annotation revealed
that 90.8% of SVs were located in intergenic regions, whereas a subset was predicted to have
potentially disruptive effects, including 6,914 frameshift, 4,045 stop-gained, and 14,587 exon-loss
variants ( **Supplementary Table 4** ).

To characterize SV dynamics at the population level, we genotyped these SVs across 500
core _Capsicum_ accessions using short-read resequencing data (Liu et al., 2023). Given the
challenges of short-read SV genotyping, particularly in repetitive and structurally complex regions,
we evaluated genotyping performance using 35 accessions for which both chromosome-scale
assemblies and short-read resequencing data were available. Using assembly-derived SV
genotypes as a benchmark, 961,086 insertions and 772,159 deletions achieved F1 scores &gt;0.7
( **Supplementary Table 2** ) and were retained as a high-confidence SV set. Allele frequencies
estimated from short-read genotyping for these SVs were highly concordant with those derived
from genome assemblies ( **Fig. 3a,b** ). Furthermore, phylogenetic relationships among the 500 core
accessions inferred from these high-confidence SVs ( **Supplementary Fig. 6** ) were largely
consistent with those inferred from genome-wide SNPs (Liu et al., 2023). We therefore used this
high-confidence SV set for subsequent population-level analyses.

**Evolutionary signatures and domestication-related patterns of SVs**
Population-scale SV genotyping revealed substantial variation in SV abundance among _Capsicum_
species and breeding groups ( **Supplementary Fig. 7** and **Supplementary Table 5** ). To investigate
patterns of SV differentiation during pepper domestication, we compared SV allele frequencies
along two independent domestication trajectories: _C. baccatum_ var. _baccatum_ to _C. baccatum_ var.
_pendulum_, and _C. annuum_ var. _glabriusculum_ to _C. annuum_ var. _annuum_ . Using a false discovery
rate threshold of &lt;0.001 and a &gt;twofold difference in allele frequency, we identified 27,035 and
37,908 differentiated SVs in these two comparisons, respectively, revealing extensive shifts in SV
frequencies between wild and domesticated groups. These differentiated SVs overlapped 4,106
and 5,648 genes, respectively ( **Supplementary Table 6** ). Notably, genes associated with
differentiated SVs in both domestication trajectories included _ACS1_ ( _Caa02_01g01760_ ) and _CCR_
( _Caa02_01g08820_ ). _ACS1_ encodes 4-coumarate:CoA ligase (4CL), which catalyzes the formation
of 4-coumaroyl-CoA, a key intermediate in the phenylpropanoid branch of capsaicinoid


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


biosynthesis (Vergnano et al., 2026). _CCR_ encodes cinnamoyl-CoA reductase, an enzyme involved
in phenylpropanoid metabolism and the production of capsaicinoid precursors (Reddy et al., 2014).
The occurrence of differentiated SVs associated with these genes in both domestication trajectories
highlights recurrent changes involving the phenylpropanoid pathway and suggests a potential link
between SV differentiation and capsaicinoid metabolism during pepper domestication.


**Fig. 3 Landscape and evolutionary patterns of structural variation in the** _**Capsicum**_ **super-pangenome.**

**a,b,** Concordance of allele frequencies between assembly-derived SV calls and graph-based SV genotypes

inferred from short-read sequencing for insertions ( **a** ) and deletions ( **b** ). **c,** Relationship between SV size

and derived allele count. Colors indicate the number of SVs in each bin on a logarithmic scale. The red

curve represents a fitted generalized additive model. **d,e,** Insertion ( **d** ) and deletion ( **e** ) burdens across major

_Capsicum_ groups. Numbers above brackets indicate P values for comparisons between wild and

domesticated groups (two-sided Student’s _t_ -tests). For each boxplot, the lower and upper bounds indicate

the first and third quartiles, respectively, the centerline indicates the median and the whiskers extend to 1.5×


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


interquartile range. PUB, _C. pubescens_ ; CHA, _C. chacoense_ ; BAB, _C. baccatum_ var. _baccatum_ ; BAP, _C._

_baccatum_ var. _pendulum_ ; FRU, _C. frutescens_ ; CHN, _C. chinense_ ; GLA, _C. annuum_ var. _glabriusculum_ ;

ANN, _C. annuum_ var. _annuum_ .


We next investigated the evolutionary dynamics and potential fitness consequences of SVs
by comparing derived allele frequency patterns between SNPs and SVs. We found a greater
enrichment of rare derived alleles among SVs than among SNPs ( **Supplementary Fig. 8a** ), a
pattern consistent with stronger purifying selection against SVs (Collins and Talkowski, 2025).
Derived allele count also decreased with increasing SV size ( **Fig. 3c** ), indicating stronger
evolutionary constraint on larger SVs, which may reflect their greater potential to disrupt genome

structure and function.


We then quantified SV burden across major _Capsicum_ populations as the number of derived
insertion and deletion alleles carried by each accession, normalized by the number of derived
synonymous SNP alleles. Both insertion and deletion burdens varied substantially among
populations ( **Fig. 3d,e** ). _C. chacoense_ exhibited the highest burdens, followed by _C. baccatum_ var.
_baccatum_ and _C. baccatum_ var. _pendulum_ . _C. frutescens_ and _C. annuum_ var. _glabriusculum_
displayed intermediate burdens, whereas _C. chinense_ and _C. annuum_ var. _annuum_ exhibited the
lowest burdens. Along the two domestication trajectories represented in our dataset, domesticated
populations carried significantly lower SV burdens than their corresponding wild progenitors ( **Fig.**
**3d,e** ). In contrast, nonsynonymous SNP (nSNPs) burdens were higher in domesticated populations
than in their wild counterparts ( **Supplementary Fig. 8b** ). These contrasting patterns of reduced
SV burden and elevated nSNP burden in domesticated populations may reflect the effects of
demographic history and differences in selection on these two classes of genetic variation during
domestication and subsequent breeding.

**Graph-based super-pangenome empowers association mapping of complex traits**
We generated a comprehensive phenotypic dataset comprising 130 traits across the pepper core
diversity panel, including 25 morphological and growth traits and 105 specialized metabolites—
four capsaicinoids, eight carotenoids, and 93 fruit volatile compounds ( **Fig. 4a** and
**Supplementary Tables 7-10** ). These data revealed extensive phenotypic diversity across the
germplasm and pronounced differences between wild and domesticated accessions. For instance,
fruit shape differed markedly between domesticated accessions and their wild relatives ( **Fig. 4b)**,
reflecting substantial morphological diversification during domestication. Pedicel orientation also
differed, with horizontal and downward pedicels occurring at substantially higher frequencies in
domesticated accessions than in their wild progenitors ( **Fig. 4c** ).

The four major capsaicinoids—capsaicin, dihydrocapsaicin, homodihydrocapsaicin, and
nordihydrocapsaicin—varied widely across the genus, underscoring extensive diversity in pepper
pungency ( **Fig. 4d** ). Capsaicin, a major contributor to pungency, exhibited the greatest variation
and accumulated to the highest levels in _C. chinense_ . In contrast, the other three major
capsaicinoids accumulated to the highest levels in _C. annuum_ var. _annuum_, revealing substantial
interspecific variation in capsaicinoid composition. Predominantly negative correlations were


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


observed between capsaicinoid levels and both carotenoid contents and several agronomic traits
( **Fig. 4e** ). The four major capsaicinoids were also positively correlated ( _P_ &lt; 0.05, _r_ ≥ 0.2) with 33
volatile compounds spanning diverse chemical classes, including terpenoids, branched-chain
esters, and C6 green-leaf volatiles ( **Fig. 4f** ).


**Fig. 4 A comprehensive phenome of the pepper core collection. a**, Overview of the phenotypic traits

measured in this study. **b,c**, Distribution of fruit shape ( **b** ) and pedicel orientation ( **c** ) across _Capsicum_

species and varieties. **d**, Contents of the four major capsaicinoids across the pepper core collection, with

accessions arranged according to their phylogenetic relationships and taxonomic groups. **e**, Correlations

between capsaicinoid levels, major agronomic traits, and carotenoid contents. **f**, Correlations between

capsaicinoid levels and fruit volatile compounds. Colors in **e** and **f** indicate Pearson correlation coefficients,

and asterisks indicate significant correlations ( _P_ &lt; 0.05). ANN, _C. annuum_ var. _annuum_ ; GLA, _C. annuum_

var. _glabriusculum_ ; CHN, _C. chinense_ ; FRU, _C. frutescens_ ; BAP, _C. baccatum_ var. _pendulum_ ; BAB, _C._

_baccatum_ var. _baccatum_ ; CHA, _C. chacoense_ ; PUB, _C. pubescens_ .


Integrating this phenotypic dataset with population-scale SNPs and SVs enabled
comprehensive genome-wide association studies (GWAS) across diverse agronomic and metabolic


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


traits. In total, we identified 2,831 significant associations across 107 of the 130 traits evaluated
( **Supplementary Table 11** ). Of these, 56 traits showed significant associations with SNPs only,
whereas 51 exhibited significant associations with both SNPs and SVs, demonstrating that
structural variation provides complementary information for dissecting the genetic architecture of
agronomic and metabolic diversity in _Capsicum_ .

Among the associated loci, a prominent signal on chromosome 4 (peak position:
Chr04:31,627,338; LD interval: 31.48–32.63 Mb) was associated with capsorubin content
( **Supplementary** **Fig. 9a** ). The alternative allele, detected exclusively in _C. chinense_, was
associated with significantly higher capsorubin levels than the reference allele present in _C._
_chinense_ and other species, including _C. baccatum_ var. _pendulum_ and _C. frutescens_
( **Supplementary Fig. 9b,c** ). Among the 21 annotated genes within the LD interval,
_Caa02_04g09950_, encoding an ethylene-responsive transcription factor RAP2-13-like protein
(hereafter RAP2), emerged as a strong candidate based on its substantially higher expression in
fruit than neighboring genes ( **Supplementary** **Fig. 9d** ). Functional validation using sustained
virus-induced gene silencing (VIGS) showed that silencing _RAP2_ resulted in an orange-red fruit
phenotype and markedly reduced _RAP2_ transcript abundance relative to red-ripe control fruits
( **Supplementary Fig. 9e,f** ). Targeted metabolomic profiling further revealed pronounced
reductions in the major red carotenoids capsanthin and capsorubin, together with decreases in
several esterified carotenoids, including zeaxanthin dipalmitate and violaxanthin dimyristate.
Conversely, several upstream carotenoid intermediates, including antheraxanthin, neoxanthin, and
lutein, accumulated to significantly higher levels following _RAP2_ silencing ( **Supplementary**
**Table 12** ). Together, these results support a role for _RAP2_ in regulating carotenoid metabolism and
promoting the accumulation of downstream red carotenoids in pepper fruit.

For capsaicinoids, a significant locus spanning 25.50–27.26 Mb on chromosome 11 was
associated with homodihydrocapsaicin content. This interval harbors four tandemly duplicated
enoyl-&#91;acyl-carrier-protein&#93; reductase (ENR) genes ( _Caa02_11g06430–Caa02_11g06460_ ), which
encode key enzymes involved in the fatty acid biosynthetic pathway that supplies precursors for
capsaicinoid side-chain formation ( **Supplementary** **Fig. 10a** ). Accessions carrying the alternative
allele of the lead SNP accumulated significantly higher levels of all four major capsaicinoids than
those carrying the reference allele ( **Supplementary Fig. 10b** ), suggesting that genetic variation at
this locus may contribute broadly to capsaicinoid accumulation.

Integration of SNP- and SV-based GWAS further resolved a structurally complex locus
associated with the sesquiterpene ylangene on chromosome 9 (~13–16 Mb). Whereas SNP-based
mapping identified a broad high-LD region, SV-based mapping revealed a local breakdown of LD
within this interval ( **Fig. 5a** ), highlighting its structural complexity. Comparative genomic analysis
revealed extensive copy-number variation (CNV) within a terpene synthase (TPS) gene cluster,
with tandem amplification occurring after the divergence of _C. rhomboideum_ and _C. pubescens_,
resulting in a marked expansion of TPS copy number in the remaining _Capsicum_ lineages ( **Fig.**
**5b** ). Synteny analysis further showed that the homologous genomic segment is located on
chromosome 3 in several early-diverging lineages, including _C. rhomboideum_, _C. pubescens_, _C._


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Fig. 5 Evolutionary and functional dissection of a terpene synthase gene cluster associated with**

**ylangene accumulation. a**, Regional Manhattan plot showing SNP-based (circles) and SV-based (triangles)
GWAS signals for ylangene content. Circle colors indicate linkage disequilibrium ( LD; _r_ ²) between the lead

SNP (Chr09:14,398,870) and neighboring SNPs. **b**, Synteny of the associated locus across representative

_Capsicum_ genomes. Terpene synthase (TPS) genes are highlighted in red. **c**, Ylangene levels in accessions

carrying the reference (Ref) or alternative (Alt) alleles of the lead SNP (Chr09:14,398,870; top) and lead

SV (Chr09:13,322,404; bottom). _P_ values were calculated using two-sided Wilcoxon rank-sum tests. For

each boxplot, the lower and upper bounds indicate the first and third quartiles, respectively, the centerline

indicates the median and the whiskers extend to 1.5× interquartile range. **d**, Expression profiles of TPS

genes within the associated interval during fruit development, together with predicted potentially high

&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


impact variants, including missense, splice-site and stop-gained variants.

_chacoense_, and _C. baccatum_, but on chromosome 9 in the Annuum clade, including _C._
_galapagoense_, _C. frutescens_, _C. chinense_, and _C. annuum_, consistent with an interchromosomal
translocation. Moreover, a lineage-specific inversion within this interval was identified in a subset
of _C. annuum_ accessions. Accessions carrying the alternative alleles of both the peak SNP and lead
SV accumulated significantly higher levels of ylangene ( **Fig. 5c** ). Multiple potentially functional
variants, including missense, splice-site, and stop-gain variants, were identified within genes of
the TPS cluster. Expression profiling further prioritized _Caa02_09g05220_ as the only TPS gene
consistently expressed throughout fruit development. Notably, _Caa02_09g05220_ harbored 20
missense and five splice-site variants ( **Fig. 5d** ). Together, these results reveal a highly dynamic
TPS locus shaped by copy-number expansion, large-scale structural rearrangements, and extensive
sequence variation, which may have collectively contributed to the diversification of sesquiterpene
profiles across _Capsicum_ .

**Pangenomic dissection of capsaicinoid biosynthetic genes**
To dissect the genetic architecture underlying capsaicinoid biosynthesis, we systematically
annotated potential capsaicinoid biosynthetic genes (CBGs) across the 48 _Capsicum_ genomes. We
identified 148 CBG families, substantially expanding the repertoire represented in individual
genomes (~97 per genome; range, 86-106) and revealing extensive PAV and CNV across the
pathway ( **Supplementary Table 13** ). Of these families, 54 (36.5%) were classified as core and 27
(18.2%) as soft-core, substantially higher than the corresponding genome-wide proportions of 14.3%
and 4.2%, respectively, indicating that the gene repertoire underlying capsaicinoid biosynthesis is
more broadly conserved across _Capsicum_ lineages than the genome-wide gene repertoire.
We next examined CBG expression in the placenta, the primary site of capsaicinoid biosynthesis,
at different stages of fruit development in the pungent cultivar “Zhangshugang” and two nonpungent mutants, c68 and ss97. We identified 15 CBGs that were consistently upregulated in the
pungent cultivar relative to both non-pungent mutants across all developmental stages ( **Fig. 6a** ),
representing a 28.3-fold enrichment of CBGs relative to the genome-wide background (Fisher’s
exact test, _P_ = 7.4 × 10 &#91;-18&#93; ). Notably, nine of these 15 CBGs were tandemly duplicated, representing
a 3.0-fold enrichment of tandem duplicates relative to the genome-wide background (Fisher’s
exact test, _P_ = 8.7 × 10 &#91;-4&#93; ) ( **Supplementary Table 14** ). These tandemly duplicated genes participate
in multiple key steps of the capsaicinoid biosynthetic pathway ( **Fig. 6b** ), highlighting the
contribution of tandem duplication to the expansion and diversification of the genetic repertoire
underlying pepper pungency.

Among these nine tandemly duplicated CBGs were two capsaicin synthase (CS) genes,
_Caa02_02g19970_ and _Caa02_02g19980_ (hereafter _CS-1_ and _CS-2_ ), located within the _Pun1_ locus.
_CS-_ 2 corresponds to _AT3_, previously shown to underlie the _Pun1_ locus (Stewart et al., 2005).
Comparative analysis across the 48 genomes identified six _AT3_ paralogous lineages at the _Pun1_
locus ( **Supplementary Table 15** and **Supplementary Fig. 11** ). These lineages showed contrasting
evolutionary patterns, ranging from widespread retention to partial loss or pseudogenization across


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Fig. 6 Expression and evolution of capsaicinoid biosynthetic genes** . **a**, Expression heatmap of

capsaicinoid biosynthetic genes (CBGs) in the placenta of the pungent cultivar ‘Zhangshugang’ and two

non-pungent mutants, c68 and ss97. Expression values are shown as log 2 (TPM + 1). “Y” denotes tandemly

duplicated genes, and “↑” indicates CBGs consistently upregulated in ‘Zhangshugang’ relative to both non
pungent mutants. **b**, Capsaicinoid biosynthetic pathway. Pathway steps involving tandemly duplicated

genes are marked with red stars, and genes preferentially upregulated in the pungent cultivar are highlighted

in blue. **c**, Regional synteny surrounding _pAMT4_ (red) and _pAMT5_ (blue) across _Capsicum_ genomes,

illustrating an interchromosomal rearrangement between chromosomes 3 and 9. **d**, Spatial expression

patterns of _CS-1_, _CS-2_ and _pAMT5_ in pepper fruit at 25 days after pollination (DAP).


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


_Capsicum_ . _CS-1_ and _CS-2_ belonged to the same _AT3_ lineage, but their orthologues formed two
distinct phylogenetic clades ( **Supplementary Fig. 12** ), indicating that the duplication giving rise
to this gene pair predated the diversification of the sampled _Capsicum_ lineages. The two genomes
of the non-pungent species _C. rhomboideum_ lacked intact copies of both _CS-1_ and _CS-2_, retaining
only a single pseudogenized copy. In contrast, 42 of the remaining 46 genomes retained intact
copies of both genes, whereas the other four carried partial or disruptive variants affecting one of
the two copies. Other _AT3_ lineages showed more extensive loss or pseudogenization. Expression
patterns further distinguished the _CS-1_ / _CS-2_ lineage from the other _AT3_ paralogous lineages.
Among the four lineages retaining intact genes at the _Pun1_ locus in ‘Zhudachang’, only _CS-1_ and
_CS-2_ were strongly expressed during capsaicinoid accumulation ( **Supplementary Fig. 13** ).
Moreover, _CS-2_ was consistently expressed at higher levels than _CS-1_ in the placenta of the
pungent cultivar ‘Zhangshugang’, indicating expression divergence between the duplicated copies.
Together, these patterns reveal extensive evolutionary remodeling of the _Pun1_ locus through gene
duplication, differential retention and loss, pseudogenization, and expression divergence among
_AT3_ paralogous lineages, with _CS-1_ and, particularly, _CS-2_ emerging as the predominant copies
associated with capsaicinoid biosynthesis.

In addition to _CS_, aminotransferase ( _pAMT_ ) is another key determinant of pepper pungency
(Tsurumaki and Sasanuma, 2019). Among the nine _pAMT_ copies identified in the _Capsicum_
pangenome ( **Supplementary Table 13** ), _pAMT5_ ( _Caa02_03g30090_ ) was the only member
preferentially upregulated in pungent fruit. Comparative genomic analysis revealed that _pAMT5_
was absent from the non-pungent _C. rhomboideum_ . Moreover, an interchromosomal translocation
between chromosomes 3 and 9 that distinguishes the Baccatum and Annuum clades reorganized
the genomic context surrounding _pAMT4_ and _pAMT5_ ( **Fig. 6c** ), suggesting a potential relationship
between large-scale chromosomal rearrangement and regulatory divergence at this locus.

Spatial transcriptomic analysis using a public dataset (Han et al., 2026) further showed that
_CS-1_, _CS-2_, and _pAMT5_ were specifically enriched in the subglandular parenchyma, a distinct cell
layer within the placental septum ( **Fig. 6d** ), revealing a spatially localized transcriptional program
associated with capsaicinoid biosynthesis. Together, these findings indicate that the broadly
conserved gene repertoire underlying capsaicinoid biosynthesis has been shaped by gene
duplication, structural rearrangement, and regulatory divergence, with key pathway genes
exhibiting highly localized expression within the placenta.

**Pangenome-enabled genomic prediction of complex traits in pepper**
We leveraged genetic variation captured by the pepper graph-based super-pangenome to develop
genomic prediction models for 130 phenotypes spanning agronomic traits, capsaicinoids,
carotenoids, and volatile organic compounds (VOCs). For each trait, informative genome-wide
variants were selected based on feature importance, and prediction models were constructed using
either SNP-only or combined SNP+SV marker panels. The resulting trait-specific panels
comprised only several hundred markers (mean = 769; **Supplementary Table 16** ). SVs constituted
a small proportion of the selected markers (1.1-3.6%), consistent with their lower abundance
relative to SNPs in the pangenome variation dataset. Five-fold cross-validation demonstrated


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


robust predictive performance across the 130 traits using these compact marker panels. The
SNP+SV models achieved a mean Pearson correlation of 0.78 between predicted and observed
phenotypic values, ranging from 0.53 to 0.98 across traits ( **Supplementary Table 16** ).
Incorporating SVs provided modest, trait-dependent improvements in prediction accuracy, with
SNP+SV models outperforming SNP-only models for 36 of the 130 traits. Collectively, these
results demonstrate that pangenome-derived genetic variation can support genomic prediction
across diverse agronomic and metabolic traits in pepper. The predictive performance achieved with
relatively compact marker panels indicates that informative variants can be identified from the
pangenome variation dataset for a broad range of phenotypes. Although SVs improved prediction
for a subset of traits, their overall contribution beyond SNPs was modest, suggesting that SNPs
capture much of the predictive signal, with SVs providing complementary information for specific

traits.


**Discussion**

The _Capsicum_ super-pangenome, integrating 48 high-quality genome assemblies spanning all five
domesticated taxa and their wild relatives, establishes a comprehensive genomic framework for
capturing diversity beyond that represented by a single reference genome. By coupling this
resource with population-scale genomic variation and extensive phenotypic data, our study
provides a multidimensional view of how gene-content variation, structural rearrangements, and
regulatory divergence contribute to agronomic and metabolic diversity in pepper.

The pan-NLRome and population-scale SV analyses revealed that domestication has
reshaped multiple layers of genomic diversity in _Capsicum_ . Along two independent domestication
trajectories, domesticated accessions contained fewer NLRs than their wild progenitors, consistent
with similar reductions reported in cucumber (Zhang and Deng, 2026). SVs, meanwhile, were
enriched for rare derived alleles, with derived allele frequencies decreasing with increasing SV
size, consistent with stronger purifying selection against larger structural changes. Domesticated
peppers carried lower SV burdens than their corresponding wild progenitors, a pattern also
reported in cucumber (Zhao et al., 2026), suggesting that demographic processes and selection
associated with domestication and subsequent breeding have contributed substantially to changes
in the SV landscape.

Integrating SNPs and SVs in GWAS provided a more comprehensive view of the genetic
architecture underlying complex agronomic and metabolic traits. Significant SV associations were
detected for 51 traits, demonstrating that structural variation provides complementary information
to SNP-based mapping. The structurally complex _TPS_ locus associated with ylangene exemplifies
this added resolution. Whereas SNP-based mapping identified a broad region of high linkage
disequilibrium that obscured the underlying genetic architecture, SV-based analysis revealed a
dynamic locus shaped by copy-number expansion, interchromosomal translocation, and lineagespecific inversion. In parallel, functional analysis of _RAP2_, identified from a locus associated with
capsorubin content, revealed a role for this ethylene-responsive transcription factor in carotenoid
metabolism and the accumulation of red carotenoids in pepper fruit. This finding expands the
regulatory network underlying fruit pigmentation beyond canonical carotenoid biosynthetic


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


enzymes.

Our pangenomic dissection of capsaicinoid biosynthesis further revealed extensive
evolutionary remodeling of the pathway through gene duplication, differential retention and loss,
structural rearrangement, and expression divergence. At the _Pun1_ locus, contrasting evolutionary
trajectories among _AT3_ paralogous lineages, together with the broad retention and strong
expression of _CS-1_ and _CS-2_, illustrate how gene duplication and subsequent divergence have
reshaped a key locus underlying pungency. The distinct expression pattern of _pAMT5_, together
with large-scale rearrangement of its genomic context, further highlights the potential
contributions of structural and regulatory variation to pathway evolution. Spatial transcriptomics
further revealed that key pathway genes are specifically expressed in the subglandular parenchyma
of the placental septum, revealing a highly localized transcriptional program associated with
capsaicinoid biosynthesis. These findings refine the cellular context of specialized metabolism in
pepper and extend previous models proposing that duplication and functional divergence of
ancestral metabolic genes contributed to the evolution of capsaicinoid biosynthesis (Kim et al.,
2014; Kusaka et al., 2024).
Beyond these evolutionary and functional insights, the super-pangenome provides a
foundation for translating expanded genomic diversity into pepper improvement. The high
genomic prediction accuracy achieved with relatively compact marker panels across diverse
agronomic and metabolic traits highlights the potential of pangenome-derived variation to inform
genomic selection. Collectively, our study establishes a comprehensive pangenomic framework
that illuminates the genomic consequences of _Capsicum_ domestication and the genetic basis of
metabolic diversification while providing valuable resources for genomics-assisted improvement.
Further integration of structural variation with functional genomic and evolutionary information
should enhance the identification and deployment of genetic diversity to develop resilient, highquality pepper varieties.

**Materials and Methods**

**Genome sequencing and de novo assembly**
High-molecular-weight (HMW) DNA was extracted from young, fresh leaves of 38 _Capsicum_
accessions using the cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987).
PacBio SMRT libraries were constructed from HMW DNA using the SMRTbell Express Template
Prep Kit 2.0 (PacBio, USA) and sequenced on the PacBio Sequel II platform to generate HiFi
reads. Hi-C libraries were prepared following the Proximo Hi-C Plant protocol (Phase Genomics,
USA) and sequenced on the Illumina NovaSeq 6000 platform to generate 150-bp paired-end reads.

HiFi reads from each of the 38 _Capsicum_ accessions were assembled into contigs using
Hifiasm (Cheng et al., 2021) (v0.19.6-r595) with default parameters. Hi-C reads were processed
to remove adapters and low-quality sequences using Trimmomatic (Bolger et al., 2014) (v0.39)
and aligned to the assembled contigs using BWA-MEM (Li, 2013) (v0.7.17) with default
parameters. Chromosome-scale assemblies were generated by scaffolding the contigs with Hi-C
data using Juicer (Durand et al., 2016) (v1.6) and 3D-DNA (Dudchenko et al., 2017) (v180419),
followed by manual curation. Chromosome anchoring and scaffolding were validated using


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


HiCExplorer (Wolff et al., 2022) (v3.0) and whole-genome alignments generated with MUMmer4
(Marçais et al., 2018) (v4.0.0rc1). Genome completeness was assessed using BUSCO (Simão et
al., 2015) (v5.3.2) with the embryophyta _ odb10 database. Base-level assembly accuracy was
evaluated using Merqury (Rhie et al., 2020) (v1.3) with HiFi reads. Assembly quality was further
evaluated using the LTR Assembly Index (LAI) (Ou et al., 2018) (v2.9.7).

**Repetitive sequence annotation and protein-coding gene prediction**
Repetitive sequences in the genome assemblies were annotated using EDTA (Ou et al., 2019)
(v2.0.1) with parameters ‘--sensitive 1 --anno 1’ and RepeatMasker (Tarailo-Graovac and Chen,
2009) (v4.1.5) with default parameters. Protein-coding genes were predicted from the repeatmasked genomes by integrating _ab initio_, homology-based and transcriptome-based evidence
using the MAKER pipeline (Cantarel et al., 2008) (v3.01.04). _Ab initio_ gene predictions were
performed using AUGUSTUS (Stanke et al., 2006) (v3.4.0), SNAP (Korf, 2004) (v2006-07-28)
and GeneMark-ES (Lomsadze et al., 2005) (v4.69_lic). Homology-based evidence included
protein sequences from published pepper genomes. For transcriptome-based evidence, PacBio IsoSeq long-read sequencing was performed on pooled fruit, leaf, flower and stem tissues. Iso-Seq
reads were processed to remove primer sequences using lima (v2.12.0; &#91;https://lima.how/),&#93;(https://lima.how/)
clustered and polished using the IsoSeq3 pipeline &#91;(https://github.com/ylipacbio/IsoSeq3), and&#93;(https://github.com/ylipacbio/IsoSeq3)
aligned to the corresponding genomes using minimap2 (Li, 2018) (v2.30-r1287).

To further improve gene prediction across both newly assembled and previously published
_Capsicum_ genomes, predicted gene models from all other genomes of the same _Capsicum_ species
were projected onto each target genome using Liftoff (Shumate and Salzberg, 2021) (v1.6.3) with
parameters ‘-copies -sc 0.97 -exclude_partial -s 0.97 -a 0.95 -p 50’. Original and projected gene
models were then integrated using EVidenceModeler (Haas et al., 2008) (v1.1.1) to generate the
final gene set for each genome.

Functional annotation of the predicted protein-coding genes was performed by aligning
protein sequences against the NCBI non-redundant (nr), UniProtKB/TrEMBL,
UniProtKB/SwissProt (Boeckmann et al., 2003) and InterPro (Mitchell et al., 2015) databases.
Gene Ontology (GO) terms were assigned using Blast2GO (Conesa and Götz, 2008) (v3.2.4).

**Gene-based pangenome construction**
Gene families across the 48 _Capsicum_ genomes were identified using OrthoFinder (Emms and
Kelly, 2019) (v2.5.4). Pairwise syntenic gene blocks were identified among the 48 genomes using
MCScanX (Wang et al., 2012) (v1.0.0), and synteny information was used to distinguish
paralogous genes that were not located within the same syntenic regions. Synonymous substitution
rates ( _Ks_ ) between syntenic gene pairs were calculated using MCScanX. The median _K_ s value
( _K_ s _median) for each syntenic block was used to exclude ancient duplications that predated
_Capsicum_ speciation. A maximum _K_ s _median cutoff of 0.344 was determined as the mean plus
three standard deviations of all _K_ s _median values for syntenic blocks between _C. rhomboideum_
and other _Capsicum_ species. Pangenes were classified into four categories based on their
distribution across the 48 genomes: core (present in all 48 genomes), soft-core (present in 43–47


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


genomes), shell (present in 2–42 genomes), and private (present in a single genome).

To reconstruct the species phylogeny, protein sequences from representative genomes of
different _Capsicum_ species, together with those from eggplant ( _Solanum melongena_ HQ-1315)
(Wei et al., 2020), tomato ( _S. lycopersicum_ Heinz v4.0) (Hosmani et al., 2019), and potato ( _S._
_tuberosum_ DM v8.1) (Yang et al., 2023), were clustered into orthogroups using OrthoFinder
(Emms and Kelly, 2019) (v2.5.4). A total of 5,020 single-copy orthologous genes were identified,
and the protein sequences of each orthogroup were aligned using MUSCLE (Edgar, 2022)(v5.1).
The resulting alignments were concatenated, and a maximum-likelihood species tree was
reconstructed using IQ-TREE (Minh et al., 2020)(v2.4.0). Divergence times were estimated using
the approximate-likelihood method implemented in MCMCtree in the PAML package (Yang, 2007)
(v4.10.10), with the species-tree topology fixed. Two secondary calibration constraints based on
previously reported divergence-time estimates were applied: 6.9–9.0 Ma for the divergence
between _S. tuberosum_ and _S. lycopersicum_ and 16.1–22.7 Ma for the divergence between _Solanum_
and _Capsicum_ &#91;(https://timetree.org/) (Kumar et al., 2022).&#93;(https://timetree.org/)

**Pan-NLRome construction**

NLR genes were identified from the 48 _Capsicum_ genome assemblies using NLR-Annotator
(Steuernagel et al., 2020) (v2.0). To validate the predicted NLR candidates, the hmmscan program
from HMMER (Johnson et al., 2010) (v3.0) was used to identify the conserved NB-ARC domain
(Pfam: PF00931) in the predicted protein sequences, using an E-value cutoff of 1e−3. Protein
domains were further annotated using InterProScan (Jones et al., 2014) (v5.71-102.0) with the
parameters ‘-dp -iprlookup -appl Pfam,Coils’. To improve coiled-coil (CC) domain prediction,
DeepCoil (Ludwiczak et al., 2019) was used for secondary structure validation. Following the
classification framework established in _Arabidopsis_ (Van de Weyer et al., 2019), NLR proteins
were categorized into four subclasses according to their domain composition: TNLs (containing at
least one TIR domain), CNLs (containing at least one CC domain), RNLs (containing at least one
RPW8 domain), and NLs (containing NB and LRR domain only). The predicted NLRs from the
48 genomes were clustered into orthogroups using OrthoFinder (Emms and Kelly, 2019) (v2.5.4)
to construct the pan-NLRome.

**Graph-based pangenome construction and variant genotyping**
A graph-based super-pangenome was constructed from 46 reference-quality _Capsicum_ genome
assemblies using the Minigraph-Cactus pipeline (Hickey et al., 2024), excluding two highly
divergent _C. rhomboideum_ genomes to reduce alignment ambiguity. The genome of _C. annuum_
var. _annuum_ cultivar ‘Zhudachang’ was used as the initial backbone, and the remaining assemblies
were incorporated sequentially into the graph. SVs (≥ 20 bp) were extracted from the graph.

SVs identified from the graph pangenome were genotyped in the 500 core accessions (Liu et
al., 2023) using PanGenie (Ebler et al., 2022) (v3.0.2) with cleaned short-read resequencing data.
To assess genotyping accuracy, 35 accessions for which both chromosome-scale assemblies and
short-read resequencing data were available were used as a validation dataset. Graph-based SV
genotypes inferred from short-read data were compared with assembly-derived SV calls, and


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


precision, recall, and F1 scores were calculated for each variant. SVs with F1 scores greater than
0.7 were retained for downstream population analyses. Allele frequencies inferred from graphbased short-read genotyping were further compared with assembly-derived allele frequencies to
evaluate genotyping consistency.

Previously identified high-quality SNPs from the 500 core accessions, originally called
against the ‘Zhangshugang’ genome (Liu et al., 2023), were lifted over to the ‘Zhudachang’
genome using whole-genome alignments generated with minimap2 (Li, 2018) (v2.30-r1287). SNP
coordinates were converted based on the genome-to-genome alignments, resulting in 98,175,412
high-confidence SNPs mapped to the ‘Zhudachang’ reference genome.

**Population genomic analyses**
To identify SVs differentiated between wild and domesticated populations, two independent
domestication trajectories were analyzed separately: _C. baccatum_ var. _baccatum_ versus _C._
_baccatum_ var. _pendulum_, and _C. annuum_ var. _glabriusculum_ versus _C. annuum_ var. _annuum_ . For
each SV, allele frequencies were calculated within each population, and differences between wild
and domesticated groups were evaluated using Fisher’s exact test. The resulting _P_ values were
adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) procedure
(Benjamini and Hochberg, 1995). SVs with an FDR &lt; 0.001 and a greater than twofold difference
in allele frequency were considered differentiated between wild and domesticated populations.

To infer the unfolded site frequency spectrum (SFS) and quantify genetic burden, ancestral
and derived states of SNPs and SVs were inferred using _C. pubescens_ as the reference for ancestralstate polarization. For each variant, the major allele in the _C. pubescens_ population was designated
as ancestral, whereas the alternative allele was considered derived. Unfolded SFSs were calculated
separately for SNPs and SVs based on derived allele frequencies. For each accession, SNP and SV
burdens were calculated as the number of derived nonsynonymous SNPs or SVs, respectively,
normalized by the number of derived synonymous SNP alleles.

**Plant growth and phenotyping**
A field trial involving 500 core _Capsicum_ accessions was conducted in Damao Village, Jiyang
District, Sanya, China (18.23°N, 109.52°E). Seeds were sown on September 20, 2022, and
seedlings were transplanted on October 22, 2022. The trial concluded on April 10, 2023. Plants
were grown in sandy loam soil under a 40-mesh nylon net. Morphological traits were recorded for
accessions that successfully produced fruits. A second field trial involving the same 500 accessions
was conducted at the same location during the 2023–2024 growing season for fruit metabolite
analyses. Only accessions yielding sufficient fruit material (&gt;500 g) were included in the
capsaicinoid, carotenoid, and volatile compound analyses. Fruits were washed, homogenized,
freeze-dried for 72 h, ground into powder, and passed through a 40-mesh sieve before metabolite
analysis.

**Capsaicinoid quantification**
For capsaicinoid analysis, 0.25 g of fruit powder was mixed with 15 mL of methanol/water (7:3,


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


v/v) containing 0.03% butylated hydroxytoluene (BHT; Sigma-Aldrich). The mixture was
vortexed for 1 min, ultrasonicated for 30 min, and centrifuged at 10,000 rpm for 5 min. The
supernatant was collected, diluted 200-fold, and filtered through a 0.22-μm membrane filter.
Capsaicinoids were quantified using a Thermo Scientific Ultimate 3000 UHPLC system coupled
to a TSQ Endura triple-quadrupole mass spectrometer equipped with an electrospray ionization
(ESI) source. Chromatographic separation was performed on a Waters XBridge Peptide BEH C18
column (100 mm × 2.1 mm, 3.5 μm) maintained at 35 °C, using 0.05% acetic acid in water (A)
and acetonitrile (B) as the mobile phase. The gradient increased from 50% to 80% B over 7 min,
followed by column re-equilibration. The flow rate was 0.4 mL/min, and the injection volume was
5 μL. Detection was carried out in multiple reaction monitoring (MRM) mode with the following
transitions: nordihydrocapsaicin (&#91;M+H&#93;+, 294.1→137.1, CE 10 V), capsaicin (&#91;M +H&#93;+,
306.1→137.1, CE 14 V), dihydrocapsaicin (&#91;M+H&#93;+, 308.2→137.1, CE 11 V), and
homodihydrocapsaicin (&#91;M-H&#93;-, 320.2→184.1, CE 15 V).

**Carotenoid quantification**
For carotenoid analysis, 0.2 g of pepper fruit powder was mixed with 30 mL of methanol-diethyl
ether (6:4, v/v) containing 0.03% BHT. The mixture was vortexed for 1 min, sonicated at 25 °C
for 30 min, and centrifuged at 10,000 rpm for 5 min. The supernatant was filtered through a 0.22μm membrane filter. Carotenoids were quantified using the same UHPLC system described above.
Chromatographic separation was performed on a YMC Carotenoid column (100 mm × 2.0 mm, 3
μm) maintained at 35 °C. The mobile phases consisted of 0.01% formic acid with 5 mmol/L
ammonium formate in water (A) and methyl tert-butyl ether containing 0.01% BHT (B). The
gradient was programmed as follows: 0-3 min, 10-15% B; 3-5 min, 15-80% B; 5-8 min, 80-100%
B; 8-10 min, 100% B; 10-10.1 min, 100-10% B; and 10.1-15 min, 10% B. The flow rate was 0.3
mL/min, and the injection volume was 5 μL. Detection was performed in positive ionization mode
with MRM. Optimized parameters were as follows: ESI source temperature, 325 °C; ion transfer
tube temperature, 325 °C; spray voltage, 3.6 kV; sheath gas, 30 arbitrary units (Arb); auxiliary gas,
10 Arb; and sweep gas, 0 Arb.

**Volatile profiling**
For volatile analysis, approximately 3 g of ripe pepper fruit powder was placed in a 20-mL
headspace vial, and 10 µL of 2-octanol (200 µg/mL in distilled water) was added as an internal
standard. Samples were equilibrated at 40 °C for 30 min, and volatile compounds were extracted
at 35 °C for 20 min using a solid-phase microextraction (SPME) fiber (DVB/C-WR/PDMS, 10
mm, 50/30 µm; GERSTEL GmbH &amp; Co. KG, Mülheim, Germany). The fiber was thermally
desorbed in the gas chromatograph (GC) injection port for 5 min under splitless conditions.
Volatile compounds were analyzed using an Agilent 7890A GC coupled to an Agilent 7200 QTOF
mass spectrometer (Agilent Technologies Inc., Santa Clara, CA, USA) equipped with a DB-5MS
capillary column (30 mm × 0.32 mm, 0.25 µm; Agilent Technologies). The temperature program
was as follows: 40 °C for 1 min, increased at 2 °C/min to 120 °C (held for 3 min), then at 10 °C/min
to 160 °C (held for 2 min), and finally at 20 °C/min to 240 °C. Helium (≥99.999%) was used as


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


the carrier gas at a flow rate of 1.0 mL/min. Mass spectra were acquired in electron ionization (EI)
mode at 70 eV, with the transfer line and ion source temperatures set to 250 °C and 230 °C,
respectively.

**Genome-wide association studies**

Genome-wide association studies (GWAS) were performed separately for SNPs and SVs using
EMMAX (Kang et al., 2010). The top five principal components and the Balding–Nichols (BN)
kinship matrix were included as covariates to account for population structure and genetic
relatedness. Genome-wide significance thresholds were determined based on the effective
numbers of independent SNPs and SVs, estimated separately using the Genetic Type I error
Calculator (Li et al., 2012). Candidate association regions were defined based on linkage
disequilibrium (LD; _r_ &#91;2&#93; ≥ 0.8) between the lead variant and neighboring variants.

**RNA-seq experiments and data analysis**
_C. annuum_ var. _annuum_ cultivar ‘Zhangshugang’ and two non-pungent mutant lines (c68 and ss97)
were used for RNA-seq analysis. Fruit placenta and pericarp tissues were collected at 15, 20, and
either 25 or 30 days after pollination (DAP), with three biological replicates per sample. Additional
raw RNA-seq datasets were obtained from previous studies (Liu et al., 2023; Chen et al., 2024).
Adapters and low-quality sequences were removed using Trimmomatic (Bolger et al., 2014) (v0.39)
with the parameters ‘TruSeq3-PE-2.fa:2:30:10:1:TRUE SLIDINGWINDOW:4:20 LEADING:3
TRAILING:3 MINLEN:40’. Poly(A) tails were trimmed using prinseq++ (Cantu et al., 2019)
(v1.2), and rRNA-derived reads were removed by mapping against an rRNA reference database
(Chuvochina et al., 2026). The cleaned reads were then aligned to the ‘Zhudachang’ genome using
HISAT2 (Kim et al., 2019) (v 2.2.1) with default parameters. Gene-level read counts were
generated using featureCounts (Liao et al., 2014) (v2.1.1) and normalized to transcripts per million
(TPM). Genes differentially expressed between ‘Zhangshugang’ and each of the two non-pungent
mutant lines were identified separately using DESeq2 (Love et al., 2014) (v1.52.0) based on raw
read counts, with thresholds of |log 2 (fold change)| ≥ 1 and an adjusted _P_ value &lt; 0.05.

**Virus-induced gene silencing**
VIGS was performed as previously described (Zhou et al., 2021; Liu et al., 2026), with minor
modifications. A 300-bp _RAP2_ -specific fragment, predicted to have no off-target effects using the
VIGS Tool (Fernandez-Pozo et al., 2015), was cloned into the pTRV2 vector. The resulting
construct was introduced into competent cells of _Agrobacterium tumefaciens_ strain GV3101 using
the freeze–thaw method, and positive clones were verified by colony PCR. Positive bacterial
cultures were incubated overnight in Luria–Bertani (LB) medium supplemented with 50 mg/L
rifampicin and 50 mg/L kanamycin at 28 °C with shaking at 200 rpm. Cells were harvested by
centrifugation at room temperature and resuspended in infiltration buffer (10 mM MgCl₂, 10 mM
MES, pH 5.6, and 200 μM acetosyringone) to a final OD₆₀₀ of 1.0. Equal volumes of _A. tumefaciens_
cultures harboring pTRV1 and either pTRV2- _RAP2_, empty pTRV2, or pTRV2- _PDS_ were mixed
and incubated at 28 °C in the dark for 4 h. Pepper seedlings with fully expanded cotyledons but no


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


true leaves (approximately 14 days after sowing) were subjected to water deprivation for 2 days.
The _Agrobacterium_ mixtures were then infiltrated into the abaxial side of the cotyledons using a
needleless syringe (approximately 10 μL per cotyledon) until the cotyledons became water-soaked.
Following infiltration, plants were maintained in darkness at 22 °C for 24 h and then transferred
to a controlled-environment growth chamber (22 °C day/18 °C night, 65% relative humidity, 16 h
light/8 h dark, and a light intensity of 150 μmol m⁻² s⁻¹). Mixtures containing pTRV1 and either
empty pTRV2 or pTRV2- _PDS_ were used as negative and positive controls, respectively. Each
treatment included 20 plants, and the experiment was independently repeated three times. Four
weeks after infiltration, newly developed leaves were collected for RT-qPCR analysis to access
_RAP2_ silencing efficiency. Plants with confirmed silencing were maintained until fruit ripening,
and ripe fruits were collected for RT-qPCR analysis, phenotypic characterization, and carotenoid
profiling.

**Genomic prediction**
Genomic prediction was conducted separately for each trait using either SNP-only or combined
SNP+SV variant sets. Variant filtering was performed independently for each trait using accessions
with available phenotypic data. Within each trait-specific subset, variants with a genotype missing
rate &gt;20%, minor allele count &lt;10, or minor allele frequency &lt;0.05 were excluded. Variants were
further pruned based on pairwise linkage disequilibrium (LD) using an _r_ _&#91;2&#93;_ threshold of 0.1. The
resulting genotype matrices were used for trait-specific marker selection and genomic prediction
using CropGBM (Yan et al., 2021) (v1.1.2). Prediction accuracy was evaluated using 50 repetitions

of five-fold cross-validation and measured as the Pearson correlation coefficient between observed

and predicted phenotypic values.

**Data availability**
Raw HiFi, Hi-C and RNA-seq reads have been deposited in the NCBI BioProject database under
accession number PRJNA1026847. Genome assemblies have been deposited in the NCBI
BioProject database under accession number PRJNA1007743. Genome assemblies and
annotations, and variant files in VCF format are available through the Pepper Genomics Database
&#91;(http://ted.bti.cornell.edu/ftp/pepper/graph_pan).&#93;(http://ted.bti.cornell.edu/ftp/pepper/graph_pan)

**Acknowledgements**
The authors thank Jing Han and Bosheng Li from Peking University Institute of Advanced
Agricultural Sciences for providing spatial expression data of _CS-1_, _CS-2,_ and _pAMT5_ . This
research was supported by Major International (Regional) Joint Research Project (32320103011),
Key Program of the National Natural Science Foundation of China (32130097), Yuelushan
Laboratory Breeding Program (YLS-2025-ZY01001), and Major Program of the National Natural
Science Foundation of China (32494784).

**Author contributions**

F.L., Z.F., and X.Zou designed and managed the project. F.L. and R.J. coordinated genome


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


sequencing. J.Z., X.Zhao, and H.S. performed genomic data analyses. F.Z., X.D., L.O., and L.Y.
conducted field trials and plant cultivation. Y.Li, H.X., S.Y., Z.L., and B.Y. performed phenotypic
evaluation. R.W., C.X., and B.H. conducted metabolite profiling and data analysis. Y.Liu collected
samples for transcriptome sequencing. H.Y. and Z.W. contributed to data curation. P.Z. performed
functional validation experiments. J.Z., X.Zhao, and H.S. wrote the manuscript. Z.F. revised the
manuscript.

**Competing interests**
The authors declare no competing interests.

**References**

**Benjamini, Y., and Hochberg, Y.** (1995). Controlling the false discovery rate: A practical and powerful


approach to multiple testing. _J. R. Stat. Soc. Ser. B_ **57** :289–300.

**Boeckmann, B., Bairoch, A., Apweiler, R., Blatter, M. C., Estreicher, A., Gasteiger, E., Martin, M. J.,**


**Michoud, K., O’Donovan, C., Phan, I., et al.** (2003). The SWISS-PROT protein knowledgebase and

its supplement TrEMBL in 2003. _Nucleic Acids Res._ **31** :365–370.

**Bolger, A. M., Lohse, M., and Usadel, B.** (2014). Trimmomatic: A flexible trimmer for Illumina sequence


data. _Bioinformatics_ **30** :2114–2120.

**Cantarel, B. L., Korf, I., Robb, S. M. C., Parra, G., Ross, E., Moore, B., Holt, C., Alvarado, A. S., and**


**Yandell, M.** (2008). MAKER: An easy-to-use annotation pipeline designed for emerging model

organism genomes. _Genome Res._ **18** :188–196.

**Cantu, V. A., Sadural, J., and Edwards, R.** (2019). PRINSEQ++, a multi-threaded tool for fast and


efficient quality control and preprocessing of sequencing datasets. _PeerJ Prepr._ **7** :e27553v1.

**Chen, W., Wang, X., Sun, J., Wang, X., Zhu, Z., Ayhan, D. H., Yi, S., Yan, M., Zhang, L., Meng, T.,**


**et al.** (2024). Two telomere-to-telomere gapless genomes reveal insights into _Capsicum_ evolution and

capsaicinoid biosynthesis. _Nat. Commun._ **15** :4295.

**Cheng, H., Concepcion, G. T., Feng, X., Zhang, H., and Li, H.** (2021). Haplotype-resolved de novo


assembly using phased assembly graphs with hifiasm. _Nat. Methods_ **18** :170–175.

**Chuvochina, M., Gerken, J., Frentrup, M., Sandikci, Y., Goldmann, R., Freese, H. M., Göker, M.,**


**Sikorski, J., Yarza, P., Quast, C., et al.** (2026). SILVA in 2026: a global core biodata resource for

rRNA within the DSMZ digital diversity. _Nucleic Acids Res._ **54** :D334–D341.

**Collins, R. L., and Talkowski, M. E.** (2025). Diversity and consequences of structural variation in the


human genome. _Nat. Rev. Genet._ **26** :443–462.

**Conesa, A., and Götz, S.** (2008). Blast2GO: A comprehensive suite for functional analysis in plant


genomics. _Int. J. Plant Genomics_ **2008** :619832.

**Delorean, E. E., Youngblood, R. C., Simpson, S. A., Schoonmaker, A. N., Scheffler, B. E., Rutter, W.**


**B., and Hulse-Kemp, A. M.** (2023). Representing true plant genomes: haplotype-resolved hybrid

pepper genome with trio-binning. _Front. Plant Sci._ **14** :1184112.

**Doyle, J. J., and Doyle, J. L.** (1987). A rapid DNA isolation procedure for small quantities of fresh leaf


tissue. _Phytochem. Bull._ **19** :11–15.

**Dudchenko, O., Batra, S. S., Omer, A. D., Nyquist, S. K., Hoeger, M., Durand, N. C., Shamim, M. S.,**


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Machol, I., Lander, E. S., Aiden, A. P., et al.** (2017). De novo assembly of the _Aedes aegypti_ genome

using Hi-C yields chromosome-length scaffolds. _Science_ **356** :92–95.

**Durand, N. C., Shamim, M. S., Machol, I., Rao, S. S. P., Huntley, M. H., Lander, E. S., and Aiden, E.**


**L.** (2016). Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. _Cell_

_Syst._ **3** :95–98.

**Ebler, J., Ebert, P., Clarke, W. E., Rausch, T., Audano, P. A., Houwaart, T., Mao, Y., Korbel, J. O.,**


**Eichler, E. E., Zody, M. C., et al.** (2022). Pangenome-based genome inference allows efficient and

accurate genotyping across a wide spectrum of variant classes. _Nat. Genet._ **54** :518–525.

**Edgar, R. C.** (2022). Muscle5: High-accuracy alignment ensembles enable unbiased assessments of


sequence homology and phylogeny. _Nat. Commun._ **13** :6968.

**Emms, D. M., and Kelly, S.** (2019). OrthoFinder: Phylogenetic orthology inference for comparative


genomics. _Genome Biol._ **20** :238.

**Fernandez-Pozo, N., Rosli, H. G., Martin, G. B., and Mueller, L. A.** (2015). The SGN VIGS tool: User

friendly software to design virus-induced gene silencing (VIGS) Constructs for functional genomics.

_Mol. Plant_ **8** :486–488.

**Haas, B. J., Salzberg, S. L., Zhu, W., Pertea, M., Allen, J. E., Orvis, J., White, O., Buell, C. R., and**


**Wortman, J. R.** (2008). Automated eukaryotic gene structure annotation using EVidenceModeler

and the Program to Assemble Spliced Alignments. _Genome Biol._ **9** :R7.

**Han, J., Tang, Y., Yue, Z., Ji, W., Geng, K., Yang, H., Jing, Q., Xu, J., Zhou, Z., Shi, Y., et al.** (2026).


Laminar patterning transcription factors orchestrate spatial metabolite partitioning in _Capsicum_ fruit.

_Nat. Plants_ **12** :1095–1115.

**Hernández-Pérez, T., Gómez-García, M. del R., Valverde, M. E., and Paredes-López, O.** (2020).


_Capsicum annuum_ (hot pepper): An ancient Latin-American crop with outstanding bioactive

compounds and nutraceutical potential. A review. _Compr. Rev. Food Sci. Food Saf._ **19** :2972–2993.

**Hickey, G., Monlong, J., Ebler, J., Novak, A. M., Eizenga, J. M., Gao, Y., Zook, J. M., Zimin, A. V.,**


**Wood, J. M. D., Wang, T., et al.** (2024). Pangenome graph construction from genome alignments

with Minigraph-Cactus. _Nat. Biotechnol._ **42** :663–673.

**Hosmani, P. S., Flores-Gonzalez, M., Geest, H. van de, Maumus, F., Bakker, L. V, Schijlen, E., Haarst,**


**J. van, Cordewener, J., Sanchez-Perez, G., Peters, S., et al.** (2019). An improved de novo assembly

and annotation of the tomato reference genome using single-molecule sequencing, Hi-C proximity

ligation and optical maps. _bioRxiv_ **2019** :767764.

**Hulse-Kemp, A. M., Maheshwari, S., Stoffel, K., Hill, T. A., Jaffe, D., Williams, S. R., Weisenfeld, N.,**


**Ramakrishnan, S., Kumar, V., Shah, P., et al.** (2018). Reference quality assembly of the 3.5-Gb

genome of _Capsicum annuum_ from a single linked-read library. _Hortic. Res._ **5** :4.

**Johnson, L. S., Eddy, S. R., and Portugaly, E.** (2010). Hidden Markov model speed heuristic and iterative


HMM search procedure. _BMC Bioinformatics_ **11** :431.

**Jones, P., Binns, D., Chang, H. Y., Fraser, M., Li, W., McAnulla, C., McWilliam, H., Maslen, J.,**


**Mitchell, A., Nuka, G., et al.** (2014). InterProScan 5: Genome-scale protein function classification.

_Bioinformatics_ **30** :1236–1240.

**Kang, H. M., Sul, J. H., Service, S. K., Zaitlen, N. A., Kong, S. Y., Freimer, N. B., Sabatti, C., and**


**Eskin, E.** (2010). Variance component model to account for sample structure in genome-wide


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


association studies. _Nat. Genet._ **42** :348–354.

**Katz, E.** (2019). Flavors and colors: The chili pepper in Europe. In _Transatlantic Trade and Global_


_Cultural Transfers Since 1492: More than Commodities_ (ed. Kaller, M.) and Jacob, F.), pp. 30–45.

London and New York.

**Kim, S., Park, M., Yeom, S.-I., Kim, Y.-M., Lee, J. M., Lee, H.-A., Seo, E., Choi, J., Cheong, K., Kim,**


**K.-T., et al.** (2014). Genome sequence of the hot pepper provides insights into the evolution of

pungency in _Capsicum_ species. _Nat. Genet._ **46** :270–278.

**Kim, S., Park, J., Yeom, S.-I., Kim, Y.-M., Seo, E., Kim, K.-T., Kim, M.-S., Lee, J. M., Cheong, K.,**


**Shin, H.-S., et al.** (2017). New reference genome sequences of hot pepper reveal the massive

evolution of plant disease-resistance genes by retroduplication. _Genome Biol._ **18** :210.

**Kim, D., Paggi, J. M., Park, C., Bennett, C., and Salzberg, S. L.** (2019). Graph-based genome alignment


and genotyping with HISAT2 and HISAT-genotype. _Nat. Biotechnol._ **37** :907–915.

**Korf, I.** (2004). Gene finding in novel genomes. _BMC Bioinformatics_ **5** :59.

**Kumar, S., Suleski, M., Craig, J. M., Kasprowicz, A. E., Sanderford, M., Li, M., Stecher, G., and**


**Hedges, S. B.** (2022). TimeTree 5: An expanded resource for species divergence times. _Mol. Biol._

_Evol._ **39** :msac174.

**Kusaka, H., Nakasato, S., Sano, K., Kobata, K., Ohno, S., Doi, M., and Tanaka, Y.** (2024). An


evolutionary view of vanillylamine synthase _pAMT_, a key enzyme of capsaicinoid biosynthesis

pathway in chili pepper. _Plant J._ **117** :1453–1465.

**Lee, J.-H., Venkatesh, J., Jo, J., Jang, S., Kim, G. W., Kim, J.-M., Han, K., Ro, N., Lee, H.-Y., Kwon,**


**J.-K., et al.** (2022). High-quality chromosome-scale genomes facilitate effective identification of

large structural variations in hot and sweet peppers. _Hortic. Res._ **9** :uhac210.

**Li, H.** (2013). Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Preprint


at arXiv https://arxiv.org/abs/1303.3997.

**Li, H.** (2018). Minimap2: pairwise alignment for nucleotide sequences. _Bioinformatics_ **34** :3094–3100.

**Li, M., Yeung, J. M. Y., Cherny, S. S., and Sham, P. C.** (2012). Evaluating the effective numbers of


independent tests and significant _p_ -value thresholds in commercial genotyping arrays and public

imputation reference datasets. _Hum. Genet._ **131** :747–756.

**Liao, Y., Smyth, G. K., and Shi, W.** (2014). featureCounts: An efficient general purpose program for


assigning sequence reads to genomic features. _Bioinformatics_ **30** :923–930.

**Liao, Y., Wang, J., Zhu, Z., Liu, Y., Chen, J., Zhou, Y., Liu, F., Lei, J., Gaut, B. S., Cao, B., et al.**


(2022). The 3D architecture of the pepper genome and its relationship to function and evolution. _Nat._

_Commun._ **13** :3479.

**Liu, F., Zhao, J., Sun, H., Xiong, C., Sun, X., Wang, X., Wang, Z., Jarret, R., Wang, J., Tang, B., et**


**al.** (2023). Genomes of cultivated and wild _Capsicum_ species provide insights into pepper

domestication and population differentiation. _Nat. Commun._ **14** :5487.

**Liu, Q., Tan, J. J., Yang, R., Sim, J. S. M., Li, X., Zhang, X., and Jiang, D.** (2026). Dynamic epigenetic


and transcriptional regulatory network in pepper fruit development and ripening. _Plant Cell_

**38** :koag211.

**Lomsadze, A., Ter-Hovhannisyan, V., Chernoff, Y. O., and Borodovsky, M.** (2005). Gene identification


in novel eukaryotic genomes by self-training algorithm. _Nucleic Acids Res._ **33** :6494–6506.


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Love, M. I., Huber, W., and Anders, S.** (2014). Moderated estimation of fold change and dispersion for


RNA-seq data with DESeq2. _Genome Biol._ **15** :550.

**Ludwiczak, J., Winski, A., Szczepaniak, K., Alva, V., and Dunin-Horkawicz, S.** (2019). DeepCoil - A


fast and accurate prediction of coiled-coil domains in protein sequences. _Bioinformatics_ **35** :2790–

2795.

**Marçais, G., Delcher, A. L., Phillippy, A. M., Coston, R., Salzberg, S. L., and Zimin, A.** (2018).


MUMmer4: A fast and versatile genome alignment system. _PLoS Comput. Biol._ **14** :e1005944.

**Minh, B. Q., Schmidt, H. A., Chernomor, O., Schrempf, D., Woodhams, M. D., Von Haeseler, A., and**


**Lanfear, R.** (2020). IQ-TREE 2: New models and efficient methods for phylogenetic inference in the

genomic era. _Mol. Biol. Evol._ **37** :1530–1534.

**Mitchell, A., Chang, H. Y., Daugherty, L., Fraser, M., Hunter, S., Lopez, R., McAnulla, C.,**


**McMenamin, C., Nuka, G., Pesseat, S., et al.** (2015). The InterPro protein families database: The

classification resource after 15 years. _Nucleic Acids Res._ **43** :D213–D221.

**Ou, S., Chen, J., and Jiang, N.** (2018). Assessing genome assembly quality using the LTR Assembly


Index (LAI). _Nucleic Acids Res._ **46** :e126.

**Ou, S., Su, W., Liao, Y., Chougule, K., Agda, J. R. A., Hellinga, A. J., Lugo, C. S. B., Elliott, T. A.,**


**Ware, D., Peterson, T., et al.** (2019). Benchmarking transposable element annotation methods for

creation of a streamlined, comprehensive pipeline. _Genome Biol._ **20** :275.

**Qin, C., Yu, C., Shen, Y., Fang, X., Chen, L., Min, J., Cheng, J., Zhao, S., Xu, M., Luo, Y., et al.** (2014).


Whole-genome sequencing of cultivated and wild peppers provides insights into _Capsicum_

domestication and specialization. _Proc. Natl. Acad. Sci._ **111** :5135–5140.

**Reddy, U. K., Almeida, A., Abburi, V. L., Alaparthi, S. B., Unselt, D., Hankins, G., Park, M., Choi,**


**D., and Nimmakayala, P.** (2014). Identification of gene-specific polymorphisms and association with

capsaicin pathway metabolites in _Capsicum annuum_ L. collections. _PLoS One_ **9** :e86393.

**Rhie, A., Walenz, B. P., Koren, S., and Phillippy, A. M.** (2020). Merqury: Reference-free quality,


completeness, and phasing assessment for genome assemblies. _Genome Biol._ **21** :245.

**Schreiber, M., Jayakodi, M., Stein, N., and Mascher, M.** (2024). Plant pangenomes for crop


improvement, biodiversity and evolution. _Nat. Rev. Genet._ **25** :563–577.

**Shirasawa, K., Hosokawa, M., Yasui, Y., Toyoda, A., and Isobe, S.** (2023). Chromosome-scale genome


assembly of a Japanese chili pepper landrace, _Capsicum annuum_ ‘Takanotsume.’ _DNA Res._

**30** :dsac052.

**Shumate, A., and Salzberg, S. L.** (2021). Liftoff: Accurate mapping of gene annotations. _Bioinformatics_


**37** :1639–1643.

**Simão, F. A., Waterhouse, R. M., Ioannidis, P., Kriventseva, E. V., and Zdobnov, E. M.** (2015).


BUSCO: Assessing genome assembly and annotation completeness with single-copy orthologs.

_Bioinformatics_ **31** :3210–3212.

**Stanke, M., Keller, O., Gunduz, I., Hayes, A., Waack, S., and Morgenstern, B.** (2006). AUGUSTUS:


_ab initio_ prediction of alternative transcripts. _Nucleic Acids Res._ **34** :435–439.

**Steuernagel, B., Witek, K., Krattinger, S. G., Ramirez-Gonzalez, R. H., Schoonbeek, H. J., Yu, G.,**


**Baggs, E., Witek, A. I., Yadav, I., Krasileva, K. V., et al.** (2020). The NLR-annotator tool enables

annotation of the intracellular immune receptor repertoire. _Plant Physiol._ **183** :468–482.


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**Stewart, C., Kang, B. C., Liu, K., Mazourek, M., Moore, S. L., Eun, Y. Y., Kim, B. D., Paran, I., and**


**Jahn, M. M.** (2005). The _Pun1_ gene for pungency in pepper encodes a putative acyltransferase. _Plant_

_J._ **42** :675–688.

**Sun, H., Zhang, J., Liao, S., Guo, S., Zhou, Z., Zhao, X., Wu, S., Zhao, J., Gong, G., Wang, J., et al.**


(2026). Population-level super-pangenome reveals genome evolution and empowers precision

breeding in watermelon. _Nat. Genet._ **58** :1176–1185.

**Tarailo-Graovac, M., and Chen, N.** (2009). Using RepeatMasker to identify repetitive elements in


genomic sequences. _Curr. Protoc. Bioinforma._ Chapter 4:4.10.1-4.10.14
**Tsurumaki, K., and Sasanuma, T.** (2019). Discovery of novel unfunctional _pAMT_ allele _pamt_ _&#91;10&#93;_ causing


loss of pungency in sweet bell pepper ( _Capsicum annuum_ L.). _Breed. Sci._ **69** :133–142.

**Van de Weyer, A. L., Monteiro, F., Furzer, O. J., Nishimura, M. T., Cevik, V., Witek, K., Jones, J. D.**


**G., Dangl, J. L., Weigel, D., and Bemm, F.** (2019). A species-wide inventory of NLR genes and

alleles in _Arabidopsis thaliana_ . _Cell_ **178** :1260-1272.e14.

**Vergnano, E., Martina, M., Poláček, P., Tikunov, Y., Gaccione, L., Barchi, L., Bovy, A., and Portis,**


**E.** (2026). Spicy genes: mapping quantitative genomic regions and candidate genes for capsaicinoid

and capsinoid biosynthesis in pepper. _Front. Plant Sci._ **17** :1823752.

**Wang, Y., Tang, H., Debarry, J. D., Tan, X., Li, J., Wang, X., Lee, T. H., Jin, H., Marler, B., Guo, H.,**


**et al.** (2012). MCScanX: A toolkit for detection and evolutionary analysis of gene synteny and

collinearity. _Nucleic Acids Res._ **40** :e49.

**Wang, S., Qian, Y. Q., Zhao, R. P., Chen, L. L., and Song, J. M.** (2023). Graph-based pan-genomes:


Increased opportunities in plant genomics. _J. Exp. Bot._ **74** :24–39.

**Wei, Q., Wang, J., Wang, W., Hu, T., Hu, H., and Bao, C.** (2020). A high-quality chromosome-level


genome assembly reveals genetics for important traits in eggplant. _Hortic. Res._ **7** :153.

**Wolff, J., Backofen, R., and Grüning, B.** (2022). Loop detection using Hi-C data with HiCExplorer.


_Gigascience_ **11** :giac06.

**Yan, J., Xu, Y., Cheng, Q., Jiang, S., Wang, Q., Xiao, Y., Ma, C., Yan, J., and Wang, X.** (2021).


LightGBM: accelerated genomically designed crop breeding through ensemble learning. _Genome Biol._

**22** :271.

**Yang, Z.** (2007). PAML 4: Phylogenetic analysis by maximum likelihood. _Mol. Biol. Evol._ **24** :1586–1591.

**Yang, X., Zhang, L., Guo, X., Xu, J., Zhang, K., Yang, Y., Yang, Y., Jian, Y., Dong, D., Huang, S., et**


**al.** (2023). The gap-free potato genome assembly reveals large tandem gene clusters of agronomical

importance in highly repeated genomic regions. _Mol. Plant_ **16** :314–317.

**Zhang, B., and Deng, Y.** (2026). Pan-NLR gene family analyses decipher evolutionary dynamics,


population diversity, and potential disease resistance genes in cucumber. _BMC Plant Biol._ **26** :345.

**Zhang, K., Yu, H., Zhang, L., Cao, Y., Li, X., Mei, Y., Wang, X., Zhang, Z., Li, T., Jin, Y., et al.** (2025).


Transposon proliferation drives genome architecture and regulatory evolution in wild and

domesticated peppers. _Nat. Plants_ **11** :359–375.

**Zhao, X., Yu, J., Zhang, J., Sun, H., Wu, S., Zhao, J., Zhou, Y., Hammar, S. A., Lin, Y. C., Zhang,**


**Z., et al.** (2026). Graph-based pangenome reveals structural variation dynamics during cucumber

breeding. _Nat. Genet._ **58** :643–654.

**Zhou, Y., Deng, Y., Liu, D., Wang, H., Zhang, X., Liu, T., Wang, J., Li, Y., Ou, L., Liu, F., et al.** (2021).


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.17.767367; this version posted November 19, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.17.767367)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


Promoting virus-induced gene silencing of pepper genes by a heterologous viral silencing suppressor.

_Plant Biotechnol. J._ **19** :2398–2400.

**Zhou, Y., Zhang, Z., Bao, Z., Li, H., Lyu, Y., Zan, Y., Wu, Y., Cheng, L., Fang, Y., Wu, K., et al.**


(2022). Graph pangenome captures missing heritability and empowers tomato breeding. _Nature_

**606** :527–534.


