Multi‐Environment Genomic Analysis Reveals Limited Concordance Between Sorghum Leaf Blight Incidence and Severity

Leaf blight caused by Exserohilum turcicum constrains sorghum production in semi‐arid environments. Disease incidence and severity are related, but their relationship need not imply shared resistance loci. Records from five locations in Niger and Senegal comprised 1339 paired plot observations. Field‐stratum‐adjusted accession values were estimated before rank‐based inverse‐normal transformation. The primary population contained 100 accessions evaluated in at least four locations; all 102 genotype‐matched accessions were included in a sensitivity analysis. Genome‐wide efficient mixed‐model association (GEMMA) score tests used leave‐one‐chromosome‐out kinship matrices, and trait‐level p ‐values were combined with the aggregated Cauchy association test (ACAT) and the Simes procedure. Adjusted incidence and standard severity were moderately correlated (Pearson r = 0.429); the correlation was lower after conditioning severity on incidence ( r = 0.066). Across 567,758 single‐nucleotide polymorphisms, ACAT had a genomic inflation factor of 1.014 and a minimum p ‐value of 2.64 × 10 −5 . No marker passed the Benjamini–Hochberg 10% false‐discovery‐rate criterion for either trait or either combination; likelihood‐ratio analyses also produced no discoveries at this threshold. Incidence–severity association‐profile concordance was weak (Pearson r = 0.127; Spearman ρ = 0.094). Top‐ranked overlap was 5.65‐fold at 1% and 2.24‐fold at 5% relative to the marker‐count expectation, a descriptive comparison that does not account for linkage disequilibrium. ACAT and Simes rankings were similar ( ρ = 0.895). Ten top‐ranked, genotype‐equivalence‐aware, linkage‐disequilibrium‐clumped regions are presented as exploratory candidates. Phenotypic association was therefore stronger than detectable sharing of marker‐association evidence. Limited sample size and sensitivity to phenotype definition, field composition and test choice constrain locus‐level conclusions.