The third variable region (V3) of the HIV-1 surface glycoprotein, gp120, plays a central role in the interaction of the virus envelope with the cell surface chemokine receptors, triggering membrane fusion and virus entry into human lymphocytes and macrophages. The CXCR4 and CCR5 chemokine receptors are used by "X4-tropic" and "R5-tropic" viruses, respectively. Recently, the crown of the V3 loop was shown to bear a close structural homology to the beta2-beta3 loop in the CXC and CC chemokines, the natural ligands of CXCR4 and CCR5, respectively. This homology can serve as the foundation for 3D molecular modeling of the V3 loops from primary isolates whose coreceptor usage was experimentally defined. The modeling revealed a charged "patch" on the surface of V3 that correlates with coreceptor usage. This V3 surface patch is positively charged in X4-tropic viruses and negatively charged or neutral in R5-tropic viruses, and is formed by two amino acids, at position 11 and at position 24 or 25; amino acids 11 and 24 or 11 and 25 contact each other in 3D space. Residues at positions 11 and 25 were known previously to influence coreceptor usage, and the charge of the residues at these two positions is often used to predict viral tropism. However, we found that the predictive value of using the charge of residues 11, 24, and 25 to identify X4 or R5 tropism was improved over using only the charge of residues 11 and 25. Thus, the data suggest a new " 11/24/25 rule" : a positively charged amino acid at position 11, 24, or 25 defines X4; otherwise R5. This rule gave an overall predictive value of 94% for 217 viruses whose tropism had been determined experimentally as either X4 or R5. The results have additional implications for the design of HIV therapeutics, vaccines, and strategies for monitoring disease progression.