Predictive modeling enabling antichlamydial discovery through virtual screening.

The data underlying Fig 2. (A) Compilation of compounds previously tested for antichlamydial activity. (B) An area under the receiver operating characteristic curve (ROC-AUC) plot showing the performance of the model after training. Data are displayed in Fig 2B. (C) An area under the precision-recall curve (PRC-AUC) plot showing the performance of the model after training. Data are displayed in Fig 2C. (D) Histogram of predicted antichlamydial hits from a virtual screen of the ChEMBL database. Data are displayed in Fig 2D. (E) Maximal structural similarity to known antibiotics (Tanimoto coefficients) and quantitative estimate of drug-likeness (QED) of the filtered hits. Data are displayed in Fig 2E. (F) The 174 hits from the virtual screen remaining after a series of filtering steps. (G) Potency of 25 compounds subjected to experimental testing, as measured by bulk GFP fluorescence using the screening assay protocol. Data are displayed in Fig 2F. (H) The 25 experimentally tested compounds from the virtual screen, tested at 11 concentrations in triplicate. A part of the data is displayed in Fig 2G. (I) Quantitative PCR-based validation of antichlamydial activity of selected compounds identified through virtual screening. Data are displayed in Fig 2G. (XLSX)