Purpose: To build a fast, user-friendly computational model to predict the intravitreal half-lives of drug-like compounds.
Methods: We used multivariate analysis to build intravitreal half-life models using two data sets, one with experimental data derived from both pigmented and albino rabbits and another including only data from experiments with albino rabbits.
Results: The final models had a Q(2) value of 0.65 and 0.75 for the mixed and albino rabbit models, respectively. The models performed well in predicting the intravitreal half-life of an external test set. In addition, the models are physiologically interpretable, containing mainly hydrogen bonding and lipophilicity descriptors.
Conclusion: The developed models enable reliable predictions of intravitreal half-lives for use in the early drug development stages, without the need for prior experimental data.