Structure-aware prediction of functional non-synonymous variants in a maize EMS mutant library

Ethyl methanesulfonate (EMS) mutagenesis is widely used to generate genetic variation for gene function studies. Previous studies have traditionally focused on stop-gained mutations, because these variants usually disrupt protein-coding capacity and produce readily detectable phenotypes. Thus, non-synonymous mutations with more moderate effects are often overlooked, despite their potential utility for fine-tuning traits in molecular breeding. Here, we present an updated maize EMS mutation resource, maizeEMSDBv2 (https://maizeems.qlnu.edu.cn/), generated by whole-genome and whole-exome sequencing datasets of 8,556 EMS-mutagenized B73-derived inbred lines. Among 20.23 million GC-to-AT mutations identified in this high gene-level coverage resource, 4,240,616 were located within genic regions, covering 96.34% of annotated genes across the maize genome. To enable systematic evaluation of 1,045,505 missense mutations, we developed the Maize Missense Effect Prediction platform, MaizeMEP (https://maizemeps.qlnu.edu.cn/), which integrates protein language model-based, structure-aware predictions of mutational effects (MEs) and biomolecular binding properties using SaProt and GPSite. Genome-wide analysis revealed that 20.8% and 15.4% of non-synonymous mutations showed moderate and high negative MEs, respectively, whereas 4.17% and 0.40% showed moderate and high positive MEs, respectively. To demonstrate the utility of MaizeMEP and associated tools, we examined three representative genes previously shown to regulate plant height in maize: Brachytic2 (ZmBR2), involved in auxin polar transport, and Kaurene synthase 3 (ZmKS3) and Anther ear 1 (ZmAN1), both involved in gibberellin biosynthesis. We identified three ZmBR2 missense variants, G419R, G636D, and S535F, that significantly reduced plant height while having limited impact on grain yield, supporting the feasibility of exploiting EMS-induced single-amino-acid substitutions for trait fine-tuning in maize molecular breeding.