Gossypium raimondii (D5) 'D5-3' Telomere-to-Telomere genome PKU_v1.0

We chose G. raimondii, commonly regarded as the donor of the D subgenomes of allotetraploid cottons, for genome sequencing. For de novo hybrid genome assembly, we used a total of 104 Gb (approximately 130-fold coverage) of Oxford Nanopore ultralong reads (N50 = 101 kb), 75 Gb of PacBio HiFi data (approximately 93-fold coverage) and 106 Gb of downloaded Hi-C library reads. The assembly was further improved and refined by following the T2T-Polish strategy. We manually confirmed and replaced four structural variants detected by PacBio HiFi reads with the criteria (homozygous alternate alleles with supporting read depth ranging from 10 to 200). Thus, a 776-Mb assembly was obtained with all 13 chromosomes represented as single gapless contigs (Supplementary Table 2). Compared with previous assemblies, our genome not only possessed substantially longer chromosomes but also provided 25 complete chromosomal ends with typical telomeric repeat units. Three recombinant DNA (rDNA) clusters previously shown using fluorescence in situ hybridization experiments31 to reside on the p-arms (short arms) of chromosomes D07, D09, and D12, respectively, remain unresolved. An extensive evaluation confirmed our T2T assembly (Methods), including the Hi-C contact heatmap, BUSCO genes, quality value (QV) score, completeness using short reads, the human T2T genome-verified method using PacBio HiFi reads, and a comparison of Bionano maps.

Within the current G. raimondii genome, 460 Mb (59.27%) of sequences were TEs, substantially higher than previously reported numbers. Overall, 35.2–39.6 Mb of corrected sequences (misoriented or lost regions) were recovered compared with previously updated assemblies. Our analysis filled in 79 or 174 existing genomic gaps (sequences between two contigs in previous assemblies), and corrected relatively large misorientations. Using approximately 118 Gb of Nanopore complementary DNA sequencing (cDNA-seq) and approximately 2.6 Tb of deep RNA sequencing (RNA-seq) data from ten cotton tissues, we obtained a total of 53,167 complete putative protein-coding genes, of which 49,914 were classified as non-TE-derived genes. The notably higher numbers of genes identified in this study, compared with any reported G. raimondii assemblies, emphasize the utility of using deep RNA-seq data to identify low-expressing genes and to fully annotate a given genome.

Supplementary Table 2. Comparisons of genome characteristics (upper) and number of proteincoding genes (lower) in the current G. raimondii assembly versus previous assemblies.

b. Genes that have initial start and stop codons, and genomic sequences without non-ATCG bases.
c. Genes that were supported by expression evidence or ≥ 1 other genome versions.
d. TE derived genes were identified through a blastn search against the plant TE database with the filter criteria set to an E-value threshold of 1e-10, requiring at least sequence similarity 80%, and a minimum alignment length of 100 bp.