Classical Volsurf approach was applied to a set of 70 carbapenem compounds acting as antibiotics. Antibacterial activity of Staphylococcus aureus SG 511 and Escherichia coli 078 representing Gram positive and Gram negative bacteria, respectively, was used for the analysis. The score plots obtained from principal component analysis showed clustering of compounds according to the activity and their loading plots explained the Volsurf descriptors responsible for the separation or peculiar behaviour of these compounds. Partial Least Square analysis yielded a seven component model for S. aureus with a cross-validated r2 (q2) value of 0.684 and conventional r2 value of 0.883 and for E. coli it is a six component model with cross-validated r2 (q2) value of 0.514 and conventional r2 value of 0.756. Both the PCA and PLS models were validated by an external test set of 15 compounds. All the compounds of the test set were fairly predicted with residual values less than one log unit. Comparatively activity data of S. aureus (Gram positive) was better explained than E. coli (Gram negative) by these models.