In silico evaluation of multilocus DNA barcodes for Aspergillus species identification

Accurate species-level identification of Aspergillus is clinically important because closely related species may differ in virulence and antifungal susceptibility. However, quantitative comparisons of commonly used DNA barcode loci remain limited. This study compared the discriminatory capacities of the internal transcribed spacer (ITS), β-tubulin ( benA ), and calmodulin ( CaM ) loci across selected Aspergillus reference genome assemblies using an in silico multilocus approach. Whole-genome sequences representing 180 Aspergillus species were retrieved from the NCBI RefSeq and GenBank databases. Barcode regions were extracted in silico using published primer pairs, and 57 species meeting the inclusion criteria for all three primary loci were evaluated for sequence variation and candidate-species reduction patterns. Principal coordinate analysis was used to visualize genetic relationships, while cyp51A was examined as an illustrative supplementary analysis for the selected Aspergillus flavus/A. austwickii reference-sequence pair, which remained indistinguishable across the three primary loci. PCoA revealed clustering associated with Aspergillus section classification, with the first two coordinates accounting for 20.4% and 18.2% of the variation represented by positive eigenvalues. PERMANOVA of the combined multilocus distance matrix showed that section membership accounted for 82.7% of the variation (R 2 = 0.827, pseudo-F = 17.475, P < 0.001). Reference-based candidate-species analysis using a ≥99% operational sequence identity threshold demonstrated differences in discriminatory performance among loci, while sensitivity analyses at ≥99.5%, ≥99.8%, and 100% showed that resolution was partly threshold-dependent. ITS showed the lowest discriminatory capacity. At ≥99%, benA and CaM each resolved 43 of 57 species to a single candidate, although benA yielded a smaller mean candidate-set size. The combined use of benA and CaM provided the strongest resolution, resolving 47 of 57 species to a single candidate, while performance at more stringent thresholds was marker-dependent. An illustrative cyp51A analysis revealed additional nucleotide differences between the selected A. flavus/A. austwickii reference sequences. Overall, discriminatory performance depended on the locus, targeted region, primer pair, reference database composition, and analytical approach, suggesting that in silico assessment may help prioritize informative barcode regions for subsequent sequence-based studies.