\(\ \ Y=X\alpha+P\beta+K\mu\ +\epsilon\)
where, Y is the phenotypic data, X is the marker effects, P is the population structure effects from MDS, K is the kinship effects, and ε is the residual.
Phenotypic variability explained by each significant marker was estimated by R2 values generated in MLM statistics output from TASSEL software. Further, the Bonferroni-corrected threshold was determined for each association analysis using 1/N (α = 0.05), where N is the number of markers tested, and quantile-quantile (QQ) plots were utilized to examine model fitness for the flower sex trait in each family.

Results

de novo genome assemblies of seven accessions in the Vitis genus

The cultivated grape (Vitis vinifera L. subsp. vinifera) was domesticated from V. vinifera subsp. sylvestris around ~6000-8000 years ago\cite{mcgovern2003ancient,Myles_2011}. Wild species in Vitis are widely used as genetic donors to improve cultivated lines in abiotic stress tolerance and disease resistance\cite{ruel2006resistance,Feechan2013,zhang2012cold}. To construct a representative core-genome for the genus Vitis and assist the breeding practice in grape, we selected two accessions from wild species, two accessions of wild species interspecific hybrids, two accessions of interspecific hybrid grape cultivars (crossing of wild and domesticated species) and three accessions of widely cultivated modern cultivars of Vitis vinifera subsp.  vinifera (Figure 1a). The genomes of Sultanina (a seedless table grape)\cite{Patel_2018} and Cabernet Sauvignon\cite{Minio_2017} were downloaded from the public database. We de novo assembled seven genomes using 10x ChromiumTM technology. A total of 39-fold to 66-fold raw sequences were collected and assembled with Supernova Assembler (version 2.0.0). The contig N50 ranged from 43.9 to 56.86 kb, and the scaffold N50 ranged from 278 kb to 2.1 Mb. For all nine genomes, more than 90% of BUSCO genes were covered completely, indicating that the assemblies covered the gene space well (Figure 1b). Additional duplicated complete BUSCOs were observed in Cabernet Sauvignon, which might due to the assembly including both haplotypes. While two locally phased pseudohaplotypes were generated by Supernova 2.0, the average distance of the adjacent SNPs between two pseudo-haplotypes ranged from 371 to 406 bp. Since the sequence similarity between these two haploblocks was high, only the pseudohap1 assembly was used to represent the genome in the downstream analysis. The molecular length, effective depth, and assembly statistics for each genome are shown in Supplementary Table 1.