The highly transformable maize inbred line LH244 represents an attractive model for gene discovery and genome engineering. However, the lack of a high‐quality genome assembly has limited its utility in functional genomics research. Here, we present a 2.29 Gb near‐complete assembly of the LH244 maize genome, with an overall base accuracy of 99.998%. Except for five gaps associated with super‐long thymine–adenine–guanine (TAG) repeat arrays, all the genome sequences were assembled from telomere to telomere (T2T). Comparative analysis revealed high genetic similarity between LH244 and B73, including 80.06% genome‐wide synteny and 90.92% of genes nearly identical. The LH244 genome was also compared with the complete Mo17 genome and revealed extensive intraspecific genomic variations. A total of 14 megabase‐scale structural variations (SVs) were identified, including a 3.15 Mb insertion, harboring 95 genes, within the 45S rDNA array of LH244 but not in the Mo17 genome. In addition, there were five knob arrays, with an average size of 21.76 Mb and the longest of 38.70 Mb, only existing in the LH244 genome. Despite the substantial variation in knob abundance, knob‐6S and knob‐8L were highly conserved between LH244 and Mo17, showing strong synteny and sequence identity, as well as consistent insertion patterns of genes and transposable elements (TEs). Overall, our study provides a near‐complete reference genome of an important transformable maize germplasm, which will serve as a much‐needed resource for functional genomics and genome editing of maize.

Genome foundation models hold great potential for advancing biological research, but modeling diverse complex genomes and generating controllable genome-scale sequences remain challenging. We present DeepD, a phylogeny-informed sparse mixture-of-experts model operating at single-nucleotide resolution over a 1-million-token context while reducing active computation. Pretrained on unselected whole-genome sequences across all domains of life, DeepD learns genic and intergenic sequence features and supports broad sequence-to-function prediction across species. DeepD performs well in human protein fitness and disease variants effect prediction, such as BRCA1 and BRCA2, with particular strength in regulatory regions, such as promoter and transcription-factor binding sites. Expert routing captures exon-intron boundaries, transposable elements, and distal cis-regulatory elements. DeepD also generates experimentally validated core promoters with specified transcriptional strengths. Finally, the Context-Anchor-Body strategy enables generation of complete yeast chromosome sequences and megabase-scale Arabidopsis segments, preserving gene structure and order while introducing variations within natural ranges. Together, DeepD is an efficient genome foundation model that unifies accurate sequence-to-function prediction with controllable genome-scale generation, advancing predictive and generative design of complex genomes.

Genome foundation models hold great potential for biological research, but modeling diverse complex genomes and generating controllable genome-scale sequences remain challenging. We present DeepD, a phylogeny-informed sparse mixture-of-experts model operating at single-nucleotide resolution with a 1-million-token context and reduced active computation. Pretrained on unselected whole genomes across all domains of life, DeepD learns genic and intergenic features and supports cross-species sequence-to-function prediction. DeepD performs well in protein fitness and disease-variant effect prediction in human, including BRCA1 and BRCA2, with particular strength in promoters and transcription-factor binding sites. Expert routing captures exon–intron boundaries, transposable elements, and distal cis-regulatory elements. DeepD generates experimentally validated maize core promoters with specified strengths. Its Context–Anchor–Body strategy generates complete yeast chromosomes and megabase-scale Arabidopsis segments, preserving gene structure and order while introducing natural-range variation. DeepD thus unifies accurate prediction with controllable genome-scale generation, advancing the predictive and generative design of complex genomes.

Kernel development is a pivotal determinant of grain yield in maize (Zea mays), a cereal crop that is fundamentally important for global food security. Here, we elucidated the essential role of VIM family genes in maize endosperm development by characterizing a miniature seed9 (mn9) mutant, in which the kernel size is reduced. Through map-based cloning, we demonstrated that the mn9 phenotype is caused by the concurrent disruption of 2 homologous genes, VIMa and VIMb. Notably, VIMa is broadly expressed and follows Mendelian segregation, whereas VIMb is an endosperm-specific imprinted gene with near-exclusive paternal expression. In maize, there exists another broadly expressed nonimprinted VIM family gene, VIMc, which is closest to VIMa in terms of expression pattern and sequence similarity, and we were unable to obtain vima;vimc double mutants. Single mutants of VIMa, VIMb, or VIMc exhibited no visible phenotypes or alterations in DNA methylation levels, whereas simultaneous loss of VIMa and VIMb function induced a reduction in genome-wide CG methylation specifically in the endosperm. In addition, we found that endoreduplication is impaired in vima;vimb mutant endosperm, leading to reduced endosperm cell ploidy, decreased cell size, and consequently diminished kernel weight. In summary, our results reveal that loss of VIMa and VIMb causes endosperm-specific CG hypomethylation, defective endoreduplication, and reduced kernel size and provide a systematic characterization of VIM family functions in maize endosperm development.