Dataset DOI: 10.5061/dryad.tqjq2bwdh

Overview

This dataset contains processed data supporting the population genomic, transcriptomic, and phenotypic analyses presented in the associated manuscript, “Additive multilocus architecture underlies fruit sweetness variation in melon.” The dataset is intended to be understandable and reusable independently of the manuscript.

The dataset is organized into four directories:

• Population_genomics/: population genomic summary statistics and local analyses around the tss9.1 and tss10.1 TSS-QTL regions.
• Transcriptome/: gene-level RNA-seq read counts and WGCNA-related datasets.
• Phenotype/: TSS measurements used for WGCNA and module–trait correlation analyses.
• Variants/: VCF files containing variant calls from whole-genome resequencing of the parental lines.

Unless otherwise noted, tabular files are tab-delimited text files (.tsv). VCF files are bgzip-compressed and accompanied by tabix index files (.tbi).

Raw whole-genome resequencing and RNA-seq data are available through NCBI SRA under BioProject PRJNA1442044. Population genomic analyses were based on publicly available melon resequencing data from Liu et al. (2020).

Experimental and analytical context

The study investigated the genetic architecture of total soluble solids (TSS), used as a measure of fruit sweetness, in melon. A recombinant inbred line (RIL) population derived from the elite sweet melon cultivars ‘B2’ and ‘Pearl’ was evaluated over four cultivation years. The transcriptomic datasets comprise fruit samples from ‘B2’ and ‘Pearl’ collected at 0, 2, 15, 29, 36, 43, and 55 days after pollination (DAP), with three biological replicates per cultivar and developmental stage.

Four stable TSS QTLs were identified by multi-year QTL analysis. Subsequent sequence and transcriptomic analyses focused on tss9.1 and tss10.1. Population genomic analyses around these two regions compared improved and landrace melon groups derived from the public resequencing dataset of Liu et al. (2020).

1. Population genomic data

These data were derived from publicly available melon resequencing data from Liu et al. (2020) and processed as described in the associated manuscript.

For the population genomic analyses, 30 improved cultivars and 30 traditional landraces were selected for each of C. melo ssp. melo and ssp. agrestis, giving 60 accessions per population category and 120 accessions in total. Subsequent analyses compared the combined Improved (IMP_M + IMP_A) and Landrace (LDR_M + LDR_A) groups.

Population genomic analyses were performed using the DHL92 reference genome (CM3.6.1). Genomic coordinates in the population genomic files follow this reference assembly.

1.1 FST (population differentiation)

Files

• FST_chr09.tsv
• FST_chr10.tsv

These files contain window-based FST estimates between the Improved and Landrace groups, calculated using the Weir and Cockerham estimator implemented in BCFtools (+weir-fst-pop). The data were used to evaluate population differentiation around the TSS-QTL regions.

Columns

Parameters

• Window size: 20 kb.
• Step size: 20 kb.

1.2 XP-CLR (cross-population composite likelihood ratio)

Files

• XPCLR_chr09.tsv
• XPCLR_chr10.tsv

These files contain XP-CLR statistics calculated to detect genomic regions showing evidence of differentiated selection between the Landrace and Improved groups. The data were used to evaluate selection signals around tss9.1 and tss10.1.

Columns

Parameters

• Additional MAF filter: MAF >= 0.05.
• Window size: 20 kb.
• Step size: 1 kb.
• --minsnps 3; --maxsnps 300; --size 20000; --step 1000.

XP-CLR values were interpreted relative to genome-wide background levels.

1.3 Linkage disequilibrium (LD)

LD decay

Files

• LD_decay_chr09_improved.tsv
• LD_decay_chr09_landrace.tsv
• LD_decay_chr10_improved.tsv
• LD_decay_chr10_landrace.tsv

These files contain LD-decay data for the indicated chromosome and population group. Each row contains a physical-distance value and the corresponding mean pairwise LD measured as r². The files do not contain column headers.

File structure

The filename identifies the chromosome (chr09 or chr10) and population group (improved or landrace).

LD decay was estimated separately for the Landrace and Improved populations using PLINK. Pairwise LD was calculated using --r2, --ld-window 99999, --ld-window-kb 500, and --ld-window-r2 0. LD decay distance was defined as the physical distance at which mean pairwise r² decayed to 0.2. No LD pruning was applied to the genome-wide LD dataset.

Local LD block lengths

Files

• Local_LD_length_tss9.1_improved.tsv
• Local_LD_length_tss9.1_landrace.tsv
• Local_LD_length_tss10.1_improved.tsv
• Local_LD_length_tss10.1_landrace.tsv

These files contain local LD-block information around the indicated TSS-QTL region. LD block lengths were calculated for the regions shown in the local LD analysis. LD blocks longer than 3 kb were rare and were not shown in the corresponding figure.

Columns

The group column is present in the tss9.1 files and is not present in the tss10.1 files.

Local LD matrices

Files

• Local_LD_matrix_tss9.1_improved.tsv
• Local_LD_matrix_tss9.1_improved_matrix_label.tsv
• Local_LD_matrix_tss9.1_landrace.tsv
• Local_LD_matrix_tss9.1_landrace_matrix_label.tsv
• Local_LD_matrix_tss10.1_improved.tsv
• Local_LD_matrix_tss10.1_improved_matrix_label.tsv
• Local_LD_matrix_tss10.1_landrace.tsv
• Local_LD_matrix_tss10.1_landrace_matrix_label.tsv

The Local_LD_matrix_*.tsv files contain pairwise LD (r²) matrices used to visualize local LD structure around the candidate QTL regions in Fig. 4.

Local LD matrices were calculated in PLINK after SNP pruning using --indep-pairwise 50 5 0.8 to reduce marker redundancy for visualization. LD heatmaps were generated in R using ggplot2.

The matrix files do not contain column headers. They are square matrices in which rows and columns correspond to the SNPs listed, in the same order, in the corresponding _matrix_label.tsv file.

Matrix structure

• Each row corresponds to one SNP.
• Each column corresponds to one SNP.
• The row/column order is defined by the corresponding _matrix_label.tsv file.
• Matrix entries are pairwise LD values (r²).
• The diagonal represents self-comparisons and is expected to be 1.
• Missing matrix entries are represented as nan in the supplied files.

The corresponding _matrix_label.tsv files contain one SNP identifier per line and do not contain a column header.

SNP dataset summary

The principal SNP datasets used for population genomic analyses were:

SNPs used for Δ allele frequency estimation were located within the genomic intervals 2200–2700 kb on chromosome 9 (including tss9.1) and 0–300 kb on chromosome 10 (including tss10.1).

1.4 Allele frequency

Files

• Allele_frequency_tss9.1_improved.tsv
• Allele_frequency_tss9.1_landrace.tsv
• Allele_frequency_tss10.1_improved.tsv
• Allele_frequency_tss10.1_landrace.tsv

These files contain site-level allele-frequency summaries for SNPs in the tss9.1 and tss10.1 regions for the indicated population group. Each file contains 60 taxa.

Columns

NA indicates that the corresponding additional allele was not observed or was not applicable at that site. nan is used only in the local LD matrix files to represent missing matrix entries. TBD in the allele-frequency files indicates that the corresponding inbreeding-coefficient value was not determined in the supplied output.

2. Transcriptome data

2.1 Gene-level read counts

File

• Gene_count_matrix.tsv

This file contains gene-level RNA-seq read counts for ‘B2’ and ‘Pearl’. Raw reads were quality-filtered using fastp v0.23.4 and aligned to the Harukei-3 reference genome using HISAT2 v2.2.1. Gene-level read counts were generated using featureCounts v2.0.6 with -s 2 -p -C -Q 30 -t exon and the Harukei-3 gene annotation. Genes with fewer than 50 total reads across all samples were excluded from downstream analyses. Differential expression analysis was performed using DESeq2 v1.42.1, with each developmental stage compared against 0 DAP within each accession. DEGs were defined as FDR < 0.01 and |log2FC| >= 0.5.

The dataset includes fruit samples collected at 0, 2, 15, 29, 36, 43, and 55 DAP, with three biological replicates per cultivar and developmental stage.

The first line of the file is featureCounts program metadata. The second line contains the column headers, followed by the gene-level count data.

Columns

Sample columns follow the format Cultivar_DAP_replicate_sorted.bam. For example, B2_15D_1_sorted.bam represents ‘B2’, 15 DAP, biological replicate 1. The sample columns cover both cultivars, all seven developmental stages, and three biological replicates per stage.

2.2 WGCNA module assignments

Files

• Supplementary_Data_1_WGCNA_module_assignment_Pearl.tsv
• Supplementary_Data_2_WGCNA_module_assignment_B2.tsv

These files provide WGCNA module assignments for genes used in network construction for ‘Pearl’ and ‘B2’, respectively. WGCNA was performed separately for ‘Pearl’ and ‘B2’ using WGCNA v1.73 in R. Gene expression values were transformed using variance-stabilizing transformation (VST; blind = FALSE). Expression data from 15, 29, 36, 43, and 55 DAP were used for network construction.

Genes used for WGCNA network construction were defined as those with FDR < 0.01 in either the 43 DAP vs 15 DAP or 55 DAP vs 15 DAP comparison. No fold-change threshold was applied to this WGCNA gene-selection criterion. A signed network was constructed. Soft-thresholding powers were selected using an approximate scale-free topology criterion of R² > 0.8, resulting in β = 12 for ‘Pearl’ and β = 16 for ‘B2’. Modules were identified with a minimum module size of 30 genes and a merge cut height of 0.25. Module eigengenes were correlated with TSS using Pearson’s correlation coefficient, and P values were adjusted using the Benjamini–Hochberg method; adjusted P < 0.05 was considered statistically significant.

Columns

The module assignments correspond to the co-expression networks used in Fig. 3A.

2.3 Starch and sucrose metabolism genes

File

• Supplementary_Data_3_WGCNA_starch_sucrose_genes.tsv

This file contains gene-level information for starch and sucrose metabolism-related genes underlying Fig. 3A, including WGCNA module assignments, expression statistics, differential-expression statistics, and expression values across developmental stages.

Columns

For the sample-expression columns, [DAP] is one of 0D, 2D, 15D, 29D, 36D, 43D, or 55D, and [replicate] is biological replicate 1, 2, or 3.

For differential-expression analyses, log2 fold-change values represent expression at 43 or 55 DAP relative to 15 DAP. Differentially expressed genes (DEGs) were defined as FDR < 0.01 and |log2FC| ≥ 0.5. This DEG definition differs from the WGCNA gene-selection criterion, which was based on FDR significance without a fold-change threshold.

2.4 SWEET family genes

File

• Supplementary_Data_4_WGCNA_SWEET_genes.tsv

This file contains gene-level information for SWEET family genes underlying Fig. 3A, including Arabidopsis homolog identifiers, WGCNA module assignments, expression statistics, differential-expression statistics, and expression values across developmental stages.

Columns

The differential-expression and missing-value conventions are the same as those described for Supplementary_Data_3_WGCNA_starch_sucrose_genes.tsv.

3. Phenotypic data

TSS measurements

File

• TSS_data.tsv

This file contains total soluble solids (TSS) measurements used as phenotypic trait data for WGCNA module–trait correlation analyses.

Columns

The TSS dataset contains measurements from ‘B2’ and ‘Pearl’ across the developmental stages represented in the file. Missing values, if present, are retained as missing rather than inferred or imputed.

4. Variant call data

Files

• variants_raw.vcf.gz
• variants_raw.vcf.gz.tbi
• variants_snps_PASS.vcf.gz
• variants_snps_PASS.vcf.gz.tbi
• variants_indels_PASS.vcf.gz
• variants_indels_PASS.vcf.gz.tbi

These files contain variant calls generated from whole-genome resequencing of the parental lines ‘B2’ and ‘Pearl’. The VCF files are VCF version 4.2, bgzip-compressed, and indexed with tabix.

The two samples in the VCF files are P15-3b and P85-3a.

4.1 VCF file types

• variants_raw.vcf.gz: raw, unfiltered variant calls containing SNPs and indels before the final PASS filtering.
• variants_snps_PASS.vcf.gz: SNP variants that passed the applied GATK hard-filtering criteria.
• variants_indels_PASS.vcf.gz: indel variants that passed the applied GATK hard-filtering criteria.

4.2 VCF fields

Standard VCF columns are:

The supplied VCFs include standard genotype and read-support fields such as GT, AD, DP, GQ, MIN_DP, PGT, PID, and PL. The VCF header provides their formal definitions.

4.3 Filtering

Raw variants were hard-filtered using GATK VariantFiltration. For SNPs:

• QD < 2.0
• QUAL < 30.0
• SOR > 4.0
• FS > 60.0
• MQ < 40.0
• MQRankSum < -12.5
• ReadPosRankSum < -8.0

For insertions and deletions (InDels):

• QD < 2.0
• QUAL < 30.0
• FS > 200.0
• SOR > 10.0
• ReadPosRankSum < -20.0

Only variants labeled PASS were retained in the corresponding PASS files. For marker development, high-confidence homozygous variants with a minimum read depth of 10 were selected from the PASS variants using VCFtools v0.1.16.

The .tbi files are tabix indexes corresponding to their respective compressed VCF files and are provided to facilitate region-based queries.

Genomic coordinates and chromosome naming conventions follow the reference genome used for the variant-calling analysis.

5. Missing values and data conventions

• NA in the allele-frequency files indicates that the corresponding additional allele was not observed or was not applicable at that site.
• nan in local LD matrix files represents a missing matrix value.
• TBD in the allele-frequency files indicates that the corresponding inbreeding-coefficient value was not determined in the supplied output.
• No values were intentionally inferred or imputed to replace missing observations in the TSS dataset.
• Empty cells in WGCNA-derived datasets indicate that the corresponding statistic or module assignment was unavailable/not applicable for that cultivar or comparison. In particular, an empty WGCNA module or statistic should not be interpreted as zero.

6. Abbreviations and terminology

• B2: parental melon cultivar ‘B2’.
• Pearl: parental melon cultivar ‘Pearl’.
• DAP: days after pollination.
• TSS: total soluble solids.
• °Brix: unit used for TSS measurements.
• QTL: quantitative trait locus.
• FST: fixation index measuring population differentiation.
• XP-CLR: cross-population composite likelihood ratio.
• LD: linkage disequilibrium.
• r²: squared correlation coefficient used as the measure of pairwise LD.
• SNP: single-nucleotide polymorphism.
• VCF: variant call format.
• WGCNA: weighted gene co-expression network analysis.
• DESeq2: differential-expression analysis software used for RNA-seq data.
• DEG: differentially expressed gene.
• FDR: false discovery rate.
• log2FC: log2 fold change.
• TAIR: The Arabidopsis Information Resource.
• GATK: Genome Analysis Toolkit.
• BCFtools: utilities for processing variant call format and genomic variant data.
• SRA: Sequence Read Archive.
• NCBI: National Center for Biotechnology Information.
• VST: variance-stabilizing transformation.

7. Figure and analysis correspondence

• Fig. 3A: WGCNA-derived datasets in Transcriptome/Supplementary_Data_1_WGCNA_module_assignment_Pearl.tsv, Supplementary_Data_2_WGCNA_module_assignment_B2.tsv, Supplementary_Data_3_WGCNA_starch_sucrose_genes.tsv, and Supplementary_Data_4_WGCNA_SWEET_genes.tsv provide the gene-level information used to show the distribution of carbohydrate metabolism and SWEET family genes across co-expression modules.
• Fig. 4A: Population_genomics/LD_decay_*.tsv files provide the genome-wide LD-decay data for Improved and Landrace populations on chromosomes 9 and 10. LD decay distance was evaluated at r² = 0.2.
• Fig. 4B: Local_LD_matrix_*.tsv and corresponding _matrix_label.tsv files provide the local LD values and SNP ordering used to reconstruct the local LD visualizations. The allele-frequency files provide site-level allele-frequency information around the candidate loci.
• Fig. 4C: Local_LD_length_*.tsv files provide LD-block length information used for the LD-block length distributions.
• Fig. 4D: FST_chr*.tsv and XPCLR_chr*.tsv provide the FST and XP-CLR datasets used to evaluate population differentiation and localized selection signals.
• Phenotype/TSS_data.tsv provides TSS phenotypic values used in WGCNA module–trait correlation analyses.
• The VCF files provide the parental variant data used for variant characterization and downstream analyses.

8. References

Liu S, Gao P, Zhu Q, Zhu Z, Liu H, Wang X, Weng Y, Gao M, Luan F. 2020. Resequencing of 297 melon accessions reveals the genomic history of improvement and loci related to fruit traits in melon. Plant Biotechnology Journal.

Yano R, et al. 2020. Genome sequence and expression atlas of the melon cultivar Harukei-3. [Reference as cited in the associated manuscript.]

9. Data access

Raw sequencing data are available through the NCBI SRA under BioProject PRJNA1442044.

The processed datasets are intended to support independent reuse and reproduction of the analyses described above. Users should consult the associated manuscript for the complete experimental design and interpretation.

10. Contact

For questions regarding the data, please contact:

Katsunori TANAKA
Faculty of Agriculture and Life Science, Hirosaki University
k-tana3@hirosaki-u.ac.jp