Sweetpotato Germplasm and Genome Resource Database

SweetPotatoDB is an integrated genomics and multi-omics database for sweetpotato (Ipomoea batatas) and related Ipomoea species. The database integrates genome assemblies, iTAK transcription-factor classifications, 529 RNA-seq samples from 40 BioProjects, 330 germplasm accessions with phenotypic records, and comprehensive functional annotations. SweetPotatoDB provides a suite of analysis tools including BLAST, eQTL exploration, core InDel marker selection, gene expression heatmaps, functional enrichment, transcription-factor family comparison, OrthoFinder-based gene family analysis, synteny visualization, and a JBrowse2-powered genome browser, serving as a valuable resource for sweetpotato functional genomics and breeding research.

The Genome Browser module is powered by JBrowse2 and supports visualization of all 10 genome assemblies. Users can navigate to genomic regions by coordinate or gene ID, and toggle annotation tracks (gene models, repeats). The Sequence ID Search function allows retrieval of genomic sequences and gene models from GFF databases, and a Functional Annotation Keyword Search enables full-text search across gene descriptions.

The BLAST module supports nucleotide and protein sequence similarity searches against genome, CDS, and protein databases for all supported species. Five programs are available: blastn, blastp, blastx, tblastn, and tblastx. Results are displayed with alignment score, E-value, identity, and direct links to the Genome Browser for each hit.

The Enrichment module supports Gene Ontology (GO) and KEGG pathway over-representation analyses for user-provided gene lists. Analysis is performed using gseapy with Fisher's exact test and Benjamini–Hochberg FDR correction. Species-specific background gene sets are pre-computed from functional annotation for each reference genome. Results include enriched terms, gene counts, adjusted p-values, and interactive bar/dot chart visualizations.

The OrthoFinder module enables users to search for orthologous gene groups across all 10 Ipomoea assemblies. Gene families were inferred using OrthoFinder v2 based on all-vs-all DIAMOND protein comparisons. For each query gene, the corresponding orthogroup members and a phylogenetic gene tree are displayed, with trees rendered by the ggtree R package.

The Gene Details module provides comprehensive information for individual gene models, including genomic location, exon–intron structure, alternative transcripts, GO terms, KEGG pathways, protein domains, and homology to Arabidopsis thaliana. Gene structure diagrams are drawn dynamically from GFF databases, with JBrowse2 integration available for configured assemblies.

The Functional Annotation module provides full-text keyword search across gene function descriptions for all supported species. Annotation data are stored in PostgreSQL and include gene descriptions, GO terms, KEGG orthologs, and best-hit UniProt/SwissProt results. Search results are paginated and sortable.

The Synteny module visualizes whole-genome syntenic relationships between any pair of supported assemblies using PAF-format alignments generated by minimap2. Interactive dotplots are rendered with Plotly.js, allowing exploration of conserved syntenic blocks across haplotypes and species. All phased haplotype assemblies are supported for intra-species comparisons.

The Phenotype Trait module provides phenotypic records for 330 sweetpotato germplasm accessions covering up to 145 traits including flesh color, skin color, vine characteristics, starch and sugar content, and geographic origin. Users can search by germplasm name or national accession number and view associated phenotype photographs. Statistical summary charts are provided for flesh color, province of origin, and germplasm type distributions.

The Transcriptomics module contains three sub-sections. RNAseq Datasets displays metadata for 529 RNA-seq samples from 40 BioProjects with links to NCBI SRA, searchable by species, tissue, condition, and year. Gene Expression allows users to query expression levels for any gene across 136 sample groups (16 tissues × 21 conditions) for three reference genomes (IBA1.0, Beauregard_v4, pasi3), displayed as interactive heatmaps. Co-expression provides signed Pearson-correlation neighborhoods for Xushu 18 (IBA1.0) and Beauregard (V4), with interactive filtering, network visualization, ranked tables, and TSV export. Both TPM and FPKM units are supported in the expression browser.

The eQTL module presents 4,408 expression quantitative trait locus blocks associated with 3,646 target genes in mature sweetpotato storage roots. Coordinates use the diploid Ipomoea trifida NSP306 v3 reference. A chromosome-by-chromosome regulatory matrix distinguishes local and distant effects, while the evidence table supports Gene ID and lead-SNP search, chromosome and significance filters, pagination, and TSV export. The data comprise 2,261 local and 2,147 distant eQTLs reported by Zhang et al. (Horticulture Research, 2020).

The Markers module builds random, non-redundant primer panels from 3,206 high-specificity core InDel markers on the 15 chromosomes of the Xushu 18 IBA r1.0 A subgenome. Users can choose one or several chromosomes, set the desired total marker count, reshuffle the panel, copy primer pairs, and export the complete selection. A chromosome-length-scaled map follows the visual logic of Figure 2 in Xiao et al. (Horticulture Research, 2023): green ticks show the full marker resource and orange ticks identify the current random panel.

The Transcription Factors module compares iTAK classifications across eight sweetpotato and wild-relative assemblies. It provides 62,013 TF records in 69 families and 12,941 transcriptional-regulator records in 24 families. Users can switch between TF and TR classifications, inspect an assembly-by-family abundance matrix, click any cell to retrieve its members, search by gene or transcript identifier, and download the filtered records as TSV.

The Genotype module includes a study-level register for 20 sweetpotato association studies and a linked catalogue of 992 currently available germplasm records from three studies. Data Information supports publication-year, journal, title, and repository exploration. Germplasm supports sample ID, cultivar name, accession name, origin, and study filtering. Both tables provide pagination and filtered TSV export. Genotype and phenotype worksheets are reserved for a later data release and are not displayed in the current version.

Genome assemblies were obtained from the SweetGAINS and GT4SP projects funded by the Bill & Melinda Gates Foundation. Gene models were annotated using a combination of ab initio prediction (AUGUSTUS), homology-based evidence (protein alignments with MAKER), and RNA-seq-guided annotation (HISAT2 + StringTie). Repeat elements were identified with RepeatMasker using species-specific repeat libraries. Functional annotations were assigned using InterProScan, eggNOG-mapper v2, and BLAST against UniProt/SwissProt.

Raw RNA-seq reads from NCBI SRA were quality-trimmed using Trimmomatic (v0.36) with parameters LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36. Clean reads were aligned to reference genomes using HISAT2 (v2.1.0). Gene expression levels were quantified as TPM and FPKM using StringTie (v2.1.4). Sample metadata were manually curated for tissue type, experimental condition, and cultivar information.

Orthologous gene groups were inferred using OrthoFinder v2 (Emms & Kelly 2019). All-vs-all protein sequence comparisons were performed with DIAMOND (v2.0) in ultra-sensitive mode across 10 Ipomoea genome assemblies. Phylogenetic gene trees were constructed using FastTree2 and visualized with the ggtree R package (Yu et al. 2017).

Whole-genome pairwise alignments were performed using minimap2 (v2.24) with assembly-to-assembly preset (-x asm5). Alignment results were stored in PAF format and filtered to retain high-quality blocks (mapping quality ≥ 20, alignment length ≥ 1 kb). Interactive dotplot visualizations are rendered in the browser using Plotly.js.

GO and KEGG enrichment analyses are performed using the gseapy Python library (Fang et al. 2023). Over-representation is assessed by Fisher's exact test; p-values are adjusted using the Benjamini–Hochberg method. Terms with adjusted p-value < 0.05 are reported as significantly enriched. Species-specific background GMT files are pre-computed from genome-wide functional annotation.

SweetPotatoDB is built on the FastAPI framework (Python) with Jinja2 templating and runs on a Linux server. The backend uses PostgreSQL for functional annotations and phenotype data, and gffutils SQLite databases for gene structure queries. The web interface uses Bootstrap 5, Plotly.js for interactive heatmaps and dotplots, Chart.js 4.4.0 for summary charts, and JBrowse2 for genome visualization. The database is available without registration and is optimized for Chrome, Firefox, Edge, and Safari.