Prediction compound promiscuity

Meta Analysis of Safety Pharmacology Data Predicting Compound Promiscuity 13.2.1... 

In the remainder of this chapter, we discuss both approaches, for target and off-target prediction of compounds, as well as for the prediction of promiscuity, based in both cases on pharmaceutical safety profiling data. 

The combined score from the fingerprint and physicochemical models was the best to confirm a logical trend from lead optimization to launched drugs. Indeed, as shown in Figure 13.4, the best scored compounds were checked for the phase in which they belonged and the average scores suggested that, as a trend, compounds predicted as promiscuous are found more often in the lead optimization phase than... 

Inhibition of the hERG ion channel is firmly associated with cardiovascular toxicity in humans, and several drugs with this liability have been withdrawn. A number of studies show that basicity, lipophilicity, and the presence of aromatic rings [76] contribute to hERG binding. The 3D models of the hERG channel [77] are potentially useful to understand more subtle structure-activity relationships. In common with receptor promiscuity, both phospholipidosis and hERG inhibition are predominantly issues with lipophilic, basic compounds, and with the predictive models available, both risks should be well controlled. 

Figure 13.3 (a) Plot of selectivity score versus promiscuity score applied to MDDR. Red dots are marketed drugs. Marketed drugs clearly cluster when compared to other compounds in different drug discovery phases (green dots), (b) The predicted selectivity of compounds in different phases of the drug discovery process. 

One way to develop an in silica tool to predictive promiscuity is to apply a NB classifier for modeling, a technique that compares the frequencies of features between selective and promiscuous sets of compounds. Bayesian classification was applied in many studies and was recently compared to other machine-learning techniques [26, 27, 43, 51, 52]. 

By mining the more profiling data, we were able to improve the prediction of previously published models for promiscuous and selective compounds. 

Based on this in-house dataset, an in-silico prediction model [27] (three-layered neural network, Ghose and Crippen [28,29] descriptors) was constructed to evaluate the frequent hitter potential before compound libraries are purchased or synthesized. This model was validated with a dataset of the above-mentioned promiscuous ligands published by McGovern et al. [26], in which 25 out of 31 compounds were correctly recognized. 

The identification of chemical features that cause compounds to be either selective or promiscuous is essential for predictive safety pharmacology methods to inform medicinal chemistry decisions. Therefore an analysis using Bayesian models along with ECFP descriptors was used. By using the compound annotation from the previous step, a multicategory Bayes model can be trained. The protocol is as follows ... 

---