A corneal-PAMPA-based in silico model for predicting corneal permeability

• QSPR models were developed based on in vitro corneal permeability measurements of 189 compounds. Partial least squares regression was applied for the model building. Classical molecular descriptors and ECFP fingerprints used. Corneal membrane retention values predicted by models. The capability to predict physicochemical parameters has always been a desirable objective ophthalmic drug development. However, previous work limited cases where either diversity compounds used lacking or performance poor. Our study provides extensive quantitative structure-property relationship (QSPR) predictions. involved diverse Preliminary analysis data showed that there is no significant correlation between corneal-PAMPA (Parallel Artificial Membrane Permeability Assay) other pharmacokinetically relevant silico transport like Caco-2, jejunal blood-brain partition coefficient (logBB). Two different developed: one retention, experimental data. producing models, which contained classical combination. A complex validation protocol (including internal external validation) carried out provide robust appropriate predictions values. Both had an overall fit R 2 > 0.90, including -values not lower than 0.85 runs, quick accurate set