ABSTRACT
We describe a novel deep learning neural network method and its application to impute assay pIC50 values. Unlike conventional machine learning approaches, this method is trained on sparse bioactivity data as input, typical of that found in public and commercial databases, enabling it to learn directly from correlations between activities measured in different assays. In two case studies on public domain data sets we show that the neural network method outperforms traditional quantitative structure-activity relationship (QSAR) models and other leading approaches. Furthermore, by focusing on only the most confident predictions the accuracy is increased to R2 > 0.9 using our method, as compared to R2 = 0.44 when reporting all predictions.
Subject(s)
Deep Learning , Pharmaceutical Preparations/chemistry , Biological Assay/methods , Databases, Pharmaceutical , Drug Discovery/methods , Molecular Structure , Quantitative Structure-Activity RelationshipABSTRACT
1. The application of novel ab initio quantum mechanical methods to the states in the catalytic cycle of cytochrome P450 following the first reduction step is described. 2. A good correlation was found between the calculated energy of reduction and the experimentally determined redox potential for a range of substrate- and substrate analogue-bound systems. 3. On reduction of the haem system, the ground state of Fe remains Fe3+. On binding of a CO molecule, Fe adopts a low-spin Fe2+ state, in agreement with experiment. However, on binding of an O2 molecule, calculations indicate that the system adopts a ferric superoxide ground state, in which the Fe is in a low-spin Fe3+ state.
Subject(s)
Camphor 5-Monooxygenase/metabolism , Energy Metabolism , Camphor 5-Monooxygenase/chemistry , Computer Simulation , Heme/chemistry , Heme/metabolism , Iron/chemistry , Iron/metabolism , Models, Molecular , Oxidation-Reduction , Pseudomonas putida/enzymology , Quantum TheoryABSTRACT
The cytochrome P450 superfamily of enzymes is ubiquitous, being responsible for the metabolism of a wide range of endogenous and xenobiotic compounds. However, the detailed mechanism of the catalytic cycle of these enzymes is still not fully understood. We describe results, obtained from first principles molecular simulations, which indicate that the low-spin state of the Fe3+ ion, present in the heme moiety at the active site of a cytochrome P450 enzyme, may be stabilized by shortening of the proximal bond of the heme. Calculations indicate that a bond length of less than approximately 2.05 A between the heme Fe3+ ion and the cysteine S, which forms the proximal ligand, would result in the stabilization of the low-spin state of the Fe3+, inhibiting the progress of the P450 catalytic cycle. Our investigation uses novel first principles modeling techniques which treat the entire system quantum-mechanically.
Subject(s)
Cytochrome P-450 Enzyme System/chemistry , Heme/chemistry , Models, MolecularABSTRACT
1. We describe the application of novel ab initio quantum mechanical methods to the study of ligand interactions with cytochrome P450cam (CYP101). 2. We find that our techniques accurately describe the transition from a low-spin state to a high-spin state of the haem Fe3+ on binding of a substrate. Furthermore, our methods correctly predict that a large fraction of low-spin character is retained on binding of an inhibitor. 3. We demonstrate the use of 'computational experiments' to elucidate key features of the mechanism of interaction. This leads us to identify a new mechanism for the suppression of the low- to high-spin transition on binding of an inhibitor, namely the shortening of the bond between the Fe atom and the coordinated S atom of the cysteine axial ligand.
Subject(s)
Camphor 5-Monooxygenase/metabolism , Computer Simulation , Ligands , Models, Theoretical , Protein Binding , Quantum TheoryABSTRACT
We discuss the use of ab initio quantum mechanical methods in drug metabolism studies. These methods require only the positions and atomic numbers of the atoms to be specified and offer greater transferability than conventional molecular modeling techniques. This fact, coupled with the accuracy of our approach, permits 'computational experiments' to be performed, allowing details of reaction mechanisms to be understood. We review the application of these methods to the cytochrome P450 superfamily of enzymes. There is much interest in understanding the mechanisms of these enzymes due to their participation in a wide range of metabolic processes including drug activation/deactivation. We find that our methods accurately reproduce the low- to high-spin transition of the haem Fe on binding of a substrate. Furthermore, we identify a new mechanism for the suppression of this spin transition, namely the shortening of the bond between the Fe atom and the coordinated S atom of the cysteine axial ligand. These results indicate that ab initio molecular modeling may be usefully applied in the study of drug metabolism and that further study of intermediate states in the P450 reaction cycle would be beneficial, particularly those which are not accessible using conventional experimental approaches.