Predicting kinase inhibitors using bioactivity matrix derived informer sets.

Prediction of compounds that are active against a desired biological target is a common step in drug discovery efforts. Virtual screening methods seek some active-enriched fraction of a library for experimental testing. Where data are too scarce to train supervised learning models for compound prioritization, initial screening must provide the necessary data. Commonly, such an initial library is selected on the basis of chemical diversity by some pseudo-random process (for example, the first few plates of a larger library) or by selecting an entire smaller library. These approaches may not produce a sufficient number or diversity of actives. An alternative approach is to select an informer set of screening compounds on the basis of chemogenomic information from previous testing of compounds against a large number of targets. We compare different ways of using chemogenomic data to choose a small informer set of compounds based on previously measured bioactivity data. We develop this Informer-Based-Ranking (IBR) approach using the Published Kinase Inhibitor Sets (PKIS) as the chemogenomic data to select the informer sets. We test the informer compounds on a target that is not part of the chemogenomic data, then predict the activity of the remaining compounds based on the experimental informer data and the chemogenomic data. Through new chemical screening experiments, we demonstrate the utility of IBR strategies in a prospective test on three kinase targets not included in the PKIS.

Practical Model Selection for Prospective Virtual Screening.

Virtual (computational) high-throughput screening provides a strategy for prioritizing compounds for experimental screens, but the choice of virtual screening algorithm depends on the data set and evaluation strategy. We consider a wide range of ligand-based machine learning and docking-based approaches for virtual screening on two protein-protein interactions, PriA-SSB and RMI-FANCM, and present a strategy for choosing which algorithm is best for prospective compound prioritization. Our workflow identifies a random forest as the best algorithm for these targets over more sophisticated neural network-based models. The top 250 predictions from our selected random forest recover 37 of the 54 active compounds from a library of 22,434 new molecules assayed on PriA-SSB. We show that virtual screening methods that perform well on public data sets and synthetic benchmarks, like multi-task neural networks, may not always translate to prospective screening performance on a specific assay of interest.

Machine Learning Consensus Scoring Improves Performance Across Targets in Structure-Based Virtual Screening.

In structure-based virtual screening, compound ranking through a consensus of scores from a variety of docking programs or scoring functions, rather than ranking by scores from a single program, provides better predictive performance and reduces target performance variability. Here we compare traditional consensus scoring methods with a novel, unsupervised gradient boosting approach. We also observed increased score variation among active ligands and developed a statistical mixture model consensus score based on combining score means and variances. To evaluate performance, we used the common performance metrics ROCAUC and EF1 on 21 benchmark targets from DUD-E. Traditional consensus methods, such as taking the mean of quantile normalized docking scores, outperformed individual docking methods and are more robust to target variation. The mixture model and gradient boosting provided further improvements over the traditional consensus methods. These methods are readily applicable to new targets in academic research and overcome the potentially poor performance of using a single docking method on a new target.

Preferred binding orientations of phenacetin in CYP1A1 and CYP1A2 are associated with isoform-selective metabolism.

Human cytochromes P450 1A1 and 1A2 play important roles in drug metabolism and chemical carcinogenesis. Although these two enzymes share high sequence identity, they display different substrate specificities and inhibitor susceptibilities. In the present studies, we investigated the structural basis for these differences with phenacetin as a probe using a number of complementary approaches, such as enzyme kinetics, stoichiometric assays, NMR, and molecular modeling. Kinetic and stoichiometric analyses revealed that substrate specificity (k(cat)/K(m)) of CYP1A2 was approximately 18-fold greater than that of CYP1A1, as expected. Moreover, despite higher H2O2 production, the coupling efficiency of reducing equivalents to acetaminophen formation in CYP1A2 was tighter than that in CYP1A1. CYP1A1, in contrast to CYP1A2, displayed much higher uncoupling, producing more water. The subsequent NMR longitudinal (T1) relaxation studies with the substrate phenacetin and its product acetaminophen showed that both compounds displayed similar binding orientations within the active site of CYP1A1 and CYP1A2. However, the distance between the OCH2 protons of the ethoxy group (site of phenacetin O-deethylation) and the heme iron was 1.5 Å shorter in CYP1A2 than in CYP1A1. The NMR findings are thus consistent with our kinetic and stoichiometric results, providing a likely molecular basis for more efficient metabolism of phenacetin by CYP1A2.