Data-quality-aware in silico ITS primer benchmarking for candidate cacao fungal diagnostics separates primer performance from archive artefacts: a proof of concept.

INTRODUCTION: Public ITS archives are widely used for fungal primer design, but incomplete records and primer-derived terminal sequence can confound in silico benchmarking. We used four cacao-associated taxa as a proof of concept for an informative-site workflow that distinguishes archive artefacts from primer-template incompatibility. METHODS: We screened records for terminal sites and used trusted primer-specific GenBank metadata to flag non-independent sites before mismatch scoring. Coverage was calculated only for informative internal sites. The workflow was applied to Moniliophthora perniciosa, M. roreri, Phytophthora palmivora, and P. megakarya under conservative rules requiring no 3'-terminal mismatch, no more than two total pair mismatches, valid orientation, and a pair-specific product window. RESULTS: Among 24,816 accession × primer-pair rows, 255 were informative, 240 were eligible, and 169 were strict hits. At the accession level, terminal position alone accounted for 287 classifications, trusted primer metadata alone for one, and both for 16. ITS1/ITS2 covered 27/31 eligible M. perniciosa records and 25/26 M. roreri records; ITS1/ITS2_KYO2 covered 37/38 eligible P. palmivora records. Long ITS2-forward + ITS4 pairs were compatible when evaluable but were rarely evaluable because most records lacked large-subunit (LSU)/28S extension. The P. palmivora ITS6 site contained one observed internal sequence across eight informative windows. CONCLUSION: Informative-site filtering replaces archive-size denominators with auditable informative denominators and identifies apparent variation associated with record construction. The results prioritize candidate pairs for bench evaluation but do not establish non-target specificity, thermodynamic suitability, or laboratory performance.