Logo do repositório

Percorrer por autor "Claus, H."

A mostrar 1 - 1 de 1
  • A multinational genomic framework for predicting β-lactam resistance in Haemophilus influenzae
    Publication . Deghmane, A.E.; Asmi, M.; Abel, S.; Bajanca-Lavado, P.; Claus, H.; D'Aeth, J.; Garcia, I.; Kiedrowska, M.; Lam, T.T.; Litt, D.; Martiny, D.; Meilleur, C.; Skoczyńska, A; Tzanakaki, G.; Xirogianni, A.; Taha, M. K.
    Background: Haemophilus influenzae resistance to β-lactams is mediated by β-lactamase and amino acid alterations in penicillin-binding protein 3 encoded by ftsI gene, which shows incomplete phenotype-genotype concordance. Objectives: To develop and evaluate a genomic classification framework to predict clinically relevant β-lactam resistance categories. Methods: We analysed 5288 H. influenzae isolates collected in eight European countries and Canada. Among these, 4833 isolates had phenotypic β-lactam susceptibility that were classified into three categories: susceptible to amoxicillin (AMX), resistant to AMX but susceptible to cefotaxime and resistant to both. A 621 bp DNA fragments of the whole ftsI gene (between codons 326 and 532) were analysed to construct category-specific k-mer vocabulary and an allele classifier using Python scripts. Performance was assessed using independent allele validation and agreement with phenotypic classification was evaluated using Cohen's κ coefficient. An additional 455 isolates lacked phenotypic data and were used for external genomic application. Results: Forty-seven frequent ftsI alleles and 34 additional alleles were used for the implementation and the refinement of the vocabulary. Subsequently, 23 other alleles were used for independent allele validation and resulted in correctly predicted resistance categories for 21 alleles (accuracy 91.3%). Agreement with phenotypic classification was high (Cohen's κ 0.853; weighted κ 0.880). Application of the classifier to 455 UK isolates predicted resistance distributions consistent with those observed in phenotypically characterized datasets. Conclusions: A recurrence-filtered k-mer-based vocabulary provides a promising standardized genomic framework that complements phenotypic AST, particularly when phenotypic testing is unavailable, incomplete or heterogeneous.
    2026-09-01Artigo científico Acesso restrito Ver mais