Streptococcus pneumoniae is a leading cause of community-acquired pneumonia and gram-positive sepsis. While multiple virulence determinants have been identified, the combination of features that determines the propensity of an isolate to cause invasive pneumococcal disease (IPD) remains unknown. In this study, we determined the genetic composition of 42 invasive and 30 noninvasive clinical isolates of serotypes 6A, 6B, and 14 by comparative genomic hybridization. Comparison of the present/absent gene matrix (i.e., comparative genomic analysis [CGA]) identified a candidate core genome consisting of 1,553 genes (73% of the TIGR4 genome), 154 genes whose presence correlated with the ability to cause IPD, and 176 genes whose presence correlated with the noninvasive phenotype. Genes identified by CGA were cross-referenced with the published signature-tagged mutagenesis studies, which served to identify core and IPD-correlated genes required for in vivo passage. Among these, two pathogenicity islands, region of diversity 8a (RD8a), which encodes a neuraminidase and V-type sodium synthase, and RD10, which encodes PsrP, a protein homologous to the platelet adhesin GspB in Streptococcus gordonii, were identified. Mice infected with a PsrP mutant were delayed in the development of bacteremia and demonstrated reduced mortality versus wild-type-infected controls. Finally, the presence of seven RDs was determined to correlate with the noninvasive phenotype, a finding that suggests some RDs may contribute to asymptomatic colonization. In conclusion, RDs are unequally distributed between invasive and noninvasive isolates, RD8a and RD10 are correlated with the propensity of an isolate to cause IPD, and PsrP is required for full virulence in mice. Copyright © 2006, American Society for Microbiology. All Rights Reserved.