Michael S. Lawless

This chapter illustrates how cheminformatics can be applied to designing novel compounds that are active at the primary target and have good predicted ADMET properties. Examples of various cheminformatics techniques are illustrated in the process of designing inhibitors that inhibit both cyclooxygenase isoforms but are more potent toward COX-2. The first step in the process is to create a knowledge database of cyclooxygenase inhibitors in the public domain. This data was analyzed to find activity cliffs – small structural changes that result in drastic changes in potency. Additional cyclooxygenase potency and selectivity trends were obtained using matched molecular pair analysis. QSAR models were then developed to predict cyclooxygenase potency and selectivity. Next, computational algorithms were used to generate novel scaffolds starting from known cyclooxygenase inhibitors. Nine virtual libraries containing 240 compounds each were constructed. Predictions from the cyclooxygenase QSAR models were used to eliminate molecules with undesirable potency or selectivity. Additionally, the compounds were screened in silico for undesirable ADMET properties, e.g., low solubility, permeability, metabolic stability, or high toxicity, using a liability scoring system known as ADMET Risk™. Eight synthetic candidates were identified from this process after incorporating knowledge gained from activity cliff analysis. Four of the compounds were synthesized and tested to measure their COX-1 and COX-2 IC₅₀ values as well as several ADME properties. The best compound, SLP0020, had a COX-1 IC₅₀ of 770 nM and COX-2 IC₅₀ of 130 nM.

Background Quantitative structure-activity (QSAR) models have enormous potential for reducing drug discovery and development costs as well as the need for animal testing. Great strides have been made in estimating their overall reliability, but to fully realize that potential, researchers and regulators need to know how confident they can be in individual predictions. Results Submodels in an ensemble model which have been trained on different subsets of a shared training pool represent multiple samples of the model space, and the degree of agreement among them contains information on the reliability of ensemble predictions. For artificial neural network ensembles (ANNEs) using two different methods for determining ensemble classification – one using vote tallies and the other averaging individual network outputs – we have found that the distribution of predictions across positive vote tallies can be reasonably well-modeled as a beta binomial distribution, as can the distribution of errors. Together, these two distributions can be used to estimate the probability that a given predictive classification will be in error. Large data sets comprised of logP, Ames mutagenicity, and CYP2D6 inhibition data are used to illustrate and validate the method. The distributions of predictions and errors for the training pool accurately predicted the distribution of predictions and errors for large external validation sets, even when the number of positive and negative examples in the training pool were not balanced. Moreover, the likelihood of a given compound being prospectively misclassified as a function of the degree of consensus between networks in the ensemble could in most cases be estimated accurately from the fitted beta binomial distributions for the training pool. Conclusions Confidence in an individual predictive classification by an ensemble model can be accurately assessed by examining the distributions of predictions and errors as a function of the degree of agreement among the constituent submodels. Further, ensemble uncertainty estimation can often be improved by adjusting the voting or classification threshold based on the parameters of the error distribution. Finally, the profiles for models whose predictive uncertainty estimates are not reliable provide clues to that effect without the need for comparison to an external test set.