Unsupervised Clustering Enables Cohort-Specific Variant Recalibration in Crops

Trait discovery in crops require reliable marker datasets, yet variant discovery remains constrained by the lack of standardised, high-confidence reference resources. Publicly available plant genomic resources often rely on outdated reference genomes, heterogenous pipelines, and generic variant hard-filtering thresholds derived from human studies. Consequently, applying the Genome Analysis Toolkit (GATK) Best Practices framework, particularly machine learning-based variant quality score recalibration (VQSR), remains challenging in crops lacking suitable training resources. Here, we present an unsupervised approach for variant quality assessment and generating cohort-specific training resources, enabling integration of VQSR into established GATK workflows without predefined high-confidence training variant sets. Using Brassica napus as a model, we use the Illumina 60k Brassica SNP array to establish a reference-specific truth set for base quality score recalibration (BQSR) and independent benchmarking. Gaussian mixture model (GMM)-based clustering of multidimensional variant annotation profiles identifies high-, intermediate-, and low-quality variant classes, which provide flexible training resources without requiring labelled training data. We further validate the approach in diploid Brassica oleracea and hexaploid wheat (Triticum aestivum), representing contrasting population characteristics and genome architectures. Across all three species, GMM-based clustering consistently identifies informative quality classes and enables cohort-specific recalibration of both SNPs and insertion-deletions (INDELs). Comparison with conventional hard filtering reveals discordant pass/fail classifications for up to 46.4% of SNPs and 37.8% of INDELs across evaluated cohorts, demonstrating that the approaches capture distinct aspects of variant quality. Our results establish unsupervised variant classification as transferable strategy for generating cohort-specific VQSR resources, extending established GATK workflows to crops lacking comprehensive genomic resources.