Aims: Several virulent clones of group B streptococcus GBS are known to be associated with certain serotypes and molecular epidemiological markers. It is unclear, however, whether the clinical significance of GBS can be predicted based solely on such molecular markers. The aim of this study was to test the hypothesis that GBS virulence can be predicted by using the molecular epidemiology markers. Methods: We examined 912 human GBS isolates in which 18 distinct molecular markers including virulence-associated mobile genetic elements, polysaccharide capsule determinants, variants of a surface antigen and invasin, and antibiotic resistance-related genes were characterised using multiplex PCR based reverse line blot assay. All strains were classified in clinically relevant invasive and colonising categories. Relationships between molecular markers and clinical phenotypes were tested using statistical and machine learning analyses. Classifier performance was evaluated by the area under receiver operator characteristic curve AUC. Results: The distribution of serotypes was comparable with those in previous reports Ia, 22.1; III, 34.7; V, 17.7. From single marker analyses, only alp3 which encodes a surface protein antigen, commonly associated with serotype V showed an increased association with invasive diseases OR 2.93, p 0.0003. Molecular serotype MS II OR 10.0, p 0.0007 had a significant association with early-onset neonatal disease when compared with late-onset diseases. Predictive analysis with logistic regression and machine learning classifiers, however, only yielded weak predictive power AUC 0.560.71, stratified 10-fold cross-validation across all the subgroups. Conclusion: While some molecular epidemiological markers are important in defining GBS clusters, a definitive predictive relationship between the molecular markers and clinical outcomes may be lacking.