Soybean (*Glycine max*) is a globally crucial crop providing protein and oil. Domesticated soybean originated from wild soybean (*Glycine soja*) approximately 7000–9000 years ago in East Asia. While predominantly self-pollinating, understanding the genomic changes during domestication and improvement is vital for genomics-assisted breeding. Previous studies using SNP arrays and whole-genome resequencing (WGS) have cataloged soybean genetic variation, but an integrated analysis incorporating a substantial number of wild soybean accessions and high-coverage WGS data remained necessary for a comprehensive understanding of deleterious mutation patterns. This study aimed to address this gap by conducting high-coverage WGS on a diverse set of soybean accessions to identify domestication-selective sweeps and analyze the patterns of deleterious mutations, ultimately leveraging the results to improve the resolution of genome-wide association studies (GWAS) for important agronomic traits.

Prior research has used SNP arrays and WGS to study soybean genetic variation. However, these studies lacked the integrated analysis of high-coverage WGS data across a large and diverse set of both domesticated and wild soybean accessions, limiting a thorough understanding of the accumulation of deleterious mutations during soybean domestication. Existing data had not been fully leveraged for genotype imputation, a method that greatly enhances the resolution of GWAS analyses in other species. The study highlighted the need to address these limitations to better understand the impact of domestication and improvement on soybean's genome.

The researchers collected whole-genome sequencing (WGS) data for 833 soybean accessions, encompassing a broad geographical range, including 418 *G. max* (332 landraces and 86 improved lines), 345 *G. soja*, and 18 hybrids. After quality control, 781 accessions were retained for the analysis. The data were mapped to the soybean Williams 82 reference genome, resulting in the identification of over 10.6 million SNPs and 1.4 million indels. Population structure and diversity were assessed using PCA and FastStructure. Recombination rates and linkage disequilibrium (LD) were estimated. Domestication-selective sweeps were detected using XP-CLR, and deleterious mutations were identified using GERP scores and SIFT scores. Finally, genotype imputation was performed to enhance the resolution of GWAS for seed protein and oil content using an existing SoySNP50K dataset.

The analysis revealed approximately 7.1% fewer deleterious mutations in domesticated soybean compared to wild soybean, with an additional 1.4% reduction observed from landraces to improved lines. A total of 183 domestication-selective sweep regions were identified, showing reduced levels of deleterious alleles. The study highlighted that the selfing nature of soybean likely contributes to the reduced deleterious allele burden. Genotype imputation based on the generated high-quality SNP variation map significantly improved the resolution of GWAS for seed protein and oil content, revealing novel association signals not previously detected.

The study's findings challenge the common observation in other crops where domestication leads to an increased accumulation of deleterious mutations. The decrease in deleterious alleles in domesticated soybean is likely a result of its self-pollinating nature, leading to a more efficient purging of these alleles. The improved resolution of GWAS through genotype imputation highlights the value of this comprehensive genetic variation map for future research in soybean improvement. The identified novel association signals provide new potential targets for breeding programs focusing on oil and protein content.

This research provides a high-resolution map of soybean genetic variation, revealing a reduction in deleterious mutations during domestication, likely facilitated by self-pollination. The creation of this resource has improved the power of GWAS, opening up the possibilities for identifying new targets for improving agronomically important traits like seed protein and oil. Future work should investigate the specific functional roles of the genes within the identified selective sweep regions and further explore the impact of deleterious alleles on soybean performance.

The study primarily focused on accessions from East Asia, which might limit the generalizability of findings to soybean populations from other regions. The definition of deleterious mutations relied on computational predictions (GERP and SIFT scores), which may not perfectly capture all functionally deleterious variants. While reference bias was addressed for SIFT analysis, the impact of potential biases related to other methodological choices merits further consideration.