KPGD: A kiwifruit pangenome database for comprehensive mining of genetic diversity in the genus Actinidia

Dear Editor,

The genus Actinidia is native to China and well known as kiwifruit; it has been classified into 75 taxa, including 54 species and 21 subspecies. Since release of the first draft genome of Actinidia chinensis ‘Hongyang’ in 2013, extensive studies have made great progress in gene cloning, genetic mapping, metabolic regulation, and molecular breeding of kiwifruit. To provide a central resource for data retrieval and utilization, we successively developed two previous versions of the kiwifruit genome database: the Kiwifruit Information Resource (KIR; http://kir.atcgn.com/) and the Kiwifruit Genome Database (KGD; https://kiwifruitgenome.org/). The KIR and KGD have played crucial roles in providing the kiwifruit research community with easy access to genome sequences of the varieties Hongyang, ‘Red5,’ and ‘White,’ as well as their structural and functional annotations. However, large amounts of new omics data have been generated for kiwifruit in recent years, necessitating further updates to the KGD for efficient storage, management, analysis, and dissemination of these large datasets.

Drawing on pangenome data platforms for rice (Yu et al., 2023), Brassica napus (Cui et al., 2023), Brassica juncea (Zhang et al., 2024), and Asteraceae (Liu et al., 2025), we constructed the comprehensive Kiwifruit PanGenome Database (KPGD) (Figure 1A), which currently comprises 55 genome assemblies (including haplotype-resolved assemblies) from 33 kiwifruit accessions of 12 different species (Figure 1B; Supplemental Table 1). Among them are seven male accessions that possess the Y chromosome, providing an excellent model system for studies of sex determination (Akagi et al., 2023; Yue et al., 2024). Moreover, advances in PacBio High-Fidelity (HiFi) and Oxford Nanopore Technology (ONT) long-read sequencing have propelled genome assembly into the “telomere-to-telomere” era, revealing completely new sequences by filling remaining gaps and resolving highly repetitive regions on chromosomes (Han et al., 2023; Yue et al., 2023). The repetitive units and gene structures of the 55 genome assemblies were predicted using several custom pipelines, resulting in an average of ∼41.8% transposable element (TE) sequences and a total of 2 376 105 protein-coding genes. Genes were annotated using the nr (https://www.ncbi.nlm.nih.gov/), UniProt (https://www.uniprot.org/), TAIR (https://www.arabidopsis.org/), SGN (https://solgenomics.net/), VitisGDB (http://vitisgdb.ynau.edu.cn/), GO (https://geneontology.org/), KEGG (https://www.genome.jp/kegg/), Pfam (http://pfam.xfam.org/), and iTAK databases (http://itak.feilab.net/cgi-bin/itak/index.cgi). Syntenic blocks and orthologous genes were also identified among the 55 assemblies, revealing patterns of gene expansion and contraction in kiwifruit.

Figure 1.

Features and functions of KPGD.

(A) Architecture of KPGD.

(B) Phylogenetic tree of all the genome assemblies. X and Y indicate the sex type; cartoon graphs of kiwifruit denote different sequencing technologies: Illumina (orange), ONT (green), HiFi Sequel I (half red), HiFi Sequel II (red), and HiFi Revio (solid red).

(C) Heatmap based on the identified pan and core genes.

(D) Numbers of pan- and core-gene clusters identified with increasing numbers of assemblies.

(E) Two search options and many online tools are provided.

Because a single reference genome is insufficient to provide comprehensive information on all genotype variants, we constructed a graph-based pangenome of kiwifruit by integrating millions of genetic variants into the Hongyang v.4.0 reference genome. To ensure more accurate results, only the 48 genome assemblies produced by long-read sequencing technologies were used to build the high-quality pangenome (Figure 1B). We obtained 1 230 638 470 SNPs, 628 728 432 insertions or deletions, and 14 906 206 structural variations (SVs) across these assemblies and used them to construct the kiwifruit pangenome graph. The graph therefore includes all forms of genetic variants, enabling improvements in unbiased variant calling and discovery of more presence–absence variations compared with reference-based mapping. By clustering 2 107 205 predicted genes from the 48 assemblies, we also constructed a protein-coding gene-based pangenome for kiwifruit consisting of 85 940 non-redundant pan-gene clusters: 9398 (10.9%) core clusters and 76 542 (89.1%) dispensable clusters (Figure 1C). Simulations of pan- and core-gene cluster numbers revealed that the number of added pan-gene clusters declined slightly and the curve of core-gene clusters nearly reached a plateau when the number of assemblies was greater than or equal to 32 (Figure 1D).

We also collected 1071 transcriptomic datasets from 15 different kiwifruit species and 2 interspecific F1 hybrids by consulting the NCBI Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra). To reduce systematic bias across different studies, we used a unified and standardized procedure to recalculate gene expression values. In addition, multiple published resequencing, metabolomic, proteomic, and epigenetic datasets were gathered and integrated into KPGD for easy use.

To bridge the gaps between diverse functional annotations, two search options are provided in KPGD: basic search and advanced search (Figure 1E). The basic search option is provided on the homepage, enabling users to quickly retrieve target genes with a given gene/protein ID or name. The advanced search option offers a set of secondary navigation menus for batch querying of gene functions (e.g., GO, KEGG, and Pfam), genomic features (e.g., transcript, coding sequence, protein, upstream, and downstream), and orthologous gene families across the selected genome assemblies. Furthermore, multiple visualization applications and web tools are implemented in KPGD to facilitate knowledge transfer and data analysis for the research community; these include JBrowse, BLAST, ID converter, sequence fetching, GO/KEGG enrichment analysis, SV detection, and CRISPR design (Figure 1E). Together, all of these search pages, analysis results, and visual representations provide access to individual genes/proteins for detailed descriptions or to other useful databases for more relevant information.

Here, we present a case study illustrating the use of KPGD to identify genetic variants in candidate genes that are potentially associated with fruit color, an important agronomic trait in kiwifruit. In previous work, construction of an intra-species pangenome of A. chinensis enabled us to identify a functional SV in the promoter of a gene homologous to BCM, which can control chlorophyll accumulation during fruit ripening (Wang et al., 2024). Compared with A. chinensis, A. eriantha always has a much higher chlorophyll content in ripening fruit, resulting in a dark-green flesh phenotype (Supplemental Figure 1). Using our constructed super pangenome of the genus Actinidia, we identified 108 consistent SVs within the coding regions of functional genes between A. chinensis and A. eriantha (Supplemental Table 2). One such SV was located in Achhyv4a29g448700.t1, whose rice homolog encodes a pentatricopeptide repeat-containing protein that functions in chloroplast development (Xiao et al., 2018); Achhyv4a29g448700.t1 may therefore be a candidate gene for the mediation of fruit coloration in different kiwifruit species.

In summary, KPGD has evolved from earlier versions (KIR and KGD) to serve as a central portal for kiwifruit genomic information. Taking full advantage of newly published genome assemblies, the graph-based kiwifruit pangenome provides a framework for exploring genetic variants that may contribute to important agronomic traits and evolutionary events. Complemented by practical tools such as SV detection and CRISPR design modules, KPGD extends beyond conventional genomic databases, offering a comprehensive platform to support the kiwifruit research and breeding community. KPGD is freely available at https://kiwifruitgenome.atcgn.com/.

Data availability

Details of the case studies and analysis methods are provided in the supplemental information.

Funding

This work was supported by the National Natural Science Foundation of China (U23A20204 and 32472680), the Anhui Provincial Natural Science Foundation (2308085MC69), and the Key Scientific Research Foundation of the Education Department of Anhui Province (2024AH050452).

Acknowledgments

We thank Qing-Yong Yang from Huazhong Agricultural University and the numerous KPGD users for their valuable suggestions. No conflict of interest is declared.

Author contributions

J.Y., Y.Z., Y.L., and Z.F. conceived the project. J.Y., B.L., X. Liu, K.L., R.L., M.U.F., U.T.D., and Z.R. collected and analyzed the data. Y.Z., X. Li, Y.W., Q.W., and Y.Y. built the website. J.Y. wrote the manuscript. Y.Z., Y.L., Z.F., and S.W. revised the manuscript. All authors read and approved the final manuscript.

Published: May 13, 2025