Figure 1.

Graphical illustration of the external predictive ability of proteochemometric models for HIV-1 protease drug susceptibility. Data for one inhibitor at a time were excluded from the dataset and predicted from proteochemometric models built on the remaining data. The predicted versus measured susceptibility values for indinavir (A) and saquinavir (B) are shown. Goodness-of-fit of the models (i.e. model data) are shown as light gray symbols in panels A and B.