Predicting the target landscape of kinase inhibitors using 3D convolutional neural networks

Many therapies in clinical trials are based on single drug-single target relationships. To further extend this concept to multi-target approaches using multi-targeted drugs, we developed a machine learning pipeline to unravel the target landscape of kinase inhibitors. This pipeline, which we call 3D-KINEssence, uses a new type of protein fingerprints (3D FP) based on the structure of kinases generated through a 3D convolutional neural network (3D-CNN). These 3D-CNN kinase fingerprints were matched to molecular Morgan fingerprints to predict the targets of each respective kinase inhibitor based on available bioactivity data. The performance of the pipeline was evaluated on two test sets: a sparse drug-target set where each drug is matched in most cases to a single target and also on a densely-covered drug-target set where each drug is matched to most if not all targets. This latter set is more challenging to train, given its non-exclusive character. Our model's root-mean-square error (RMSE) based on the two datasets was 0.68 and 0.8, respectively. These results indicate that 3D FP can predict the target landscape of kinase inhibitors at around 0.8 log units of bioactivity. Our strategy can be utilized in proteochemometric or chemogenomic workflows by consolidating the target landscape of kinase inhibitors. In this publication we have set up a new method to predict the targets kinase inhibitors. This information is important since it allows to understand how the inhibition of multiple targets by an inhibitor translates to its potency. For this, we have used a convolutional neural networks (CNN) strategy that is commonly used to classify images with common but also different elements, for instance to recognize classes (predicting which animal is shown) or subtle changes (predicting which emotions are shown on a face). In this case we have used CNN models to detect patterns in 3-dimensional information from kinase structures and matched this to the bioactivity of their respective inhibitors. We show that our method has a similar performance as benchmark methods in the field. However, since recognized patterns can be made explicit, our method could help to make the underlying artificial intelligence method explainable. In the future this could lead to more optimal matching of multi-target drugs to diseases that have multiple vulnerabilities.