Integrating GWAS, transcriptomics, and machine learning enables genomic prediction of grain cadmium content in malting barley

Cadmium (Cd) contamination in arable land poses a serious threat to the production of malting barley, a cornerstone crop for the brewing industry. Excessive Cd in grains undermines malting quality by disrupting germination evenness and reducing enzymatic activity during mashing. Despite this challenge, effective predictive tools for grain Cd levels are scarce, and the genetic architecture controlling this trait remains largely elusive. Here, we employed an integrated strategy combining genome-wide association study (GWAS), RNA-seq, and machine learning across a panel of 100 barley accessions. We benchmarked five genomic prediction algorithms, performed cis-eQTL mapping, and applied multi-omics integration to nominate high-confidence candidate genes. Our GWAS uncovered 3902 Cd-associated SNPs, and the XGBoost model, trained on merely 1000 markers, achieved the highest overall predictive performance (r = 0.86). Through systematic prioritization, we identified 67 candidate metal transporter genes. Among these, HvACA8a emerged as a likely negative modulator of grain Cd loading, whereas HvAHA9 and HvYSL15 were implicated as positive regulators. Collectively, our findings establish a robust genomic prediction framework for early Cd risk evaluation in malting barley. The identified markers and candidate genes offer tangible targets for breeding low-Cd cultivars, thereby helping to safeguard the quality of raw materials destined for the malting and brewing sectors.