maize microprotein

Microproteins represent a largely unexplored coding layer in plant genomes. Here, we developed DeepMp, a hybrid deep learning framework that integrated multi-omics evidence, to systematically identify 18,338 high-confidence microproteins in maize—the most comprehensive set reported in plants. Most lacked canonical domains and originated from noncoding regions, revealing extensive hidden coding potential. Evolutionary analyses revealed rapid lineage-specific expansion, with younger microproteins evolving more rapidly and exhibiting pronounced tissue specificity. These microproteins were integrated into regulatory networks controlling organ development and nutrient storage. Population-scale analyses linked microprotein expression to kernel amino acid content, with three grain-filling–specific candidates originating from noncoding regions and exhibiting haplotype-dependent effects. CRISPR-Cas9 knockouts confirmed their roles as precise modulators of arginine, aspartate, and methionine metabolism, without causing visible changes in kernel morphology. This study provides a foundational resource and analytical framework, establishing microproteins as a functionally relevant coding layer in maize and an important source for crop improvement.