pBRICS: A Novel Fragmentation Method for Explainable Property Prediction of Drug-Like Small Molecules.

Generative artificial intelligence algorithms have shown to be successful in exploring large chemical spaces and designing novel and diverse molecules. There has been considerable interest in developing predictive models using artificial intelligence for drug-like properties, which can potentially reduce the late-stage attrition of drug candidates or predict the properties of novel AI-designed molecules. Concurrently, it is important to understand the contribution of functional groups toward these properties and modify them to obtain property-optimized lead compounds. As a result, there is an increasing interest in the development of explainable property prediction models. However, current explainable approaches are mostly atom-based, where, often, only a fraction of a fragment is shown to be significant. To address the above challenges, we have developed a novel domain-aware molecular fragmentation approach termed post-processing of BRICS (pBRICS), which can fragment small molecules into their functional groups. Multitask models were developed to predict various properties, including the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties. The fragment importance was explained using the gradient-weighted class activation mapping (Grad-CAM) approach. The method was validated on data sets of experimentally available matched molecular pairs (MMPs). The explanations from the model can be useful for medicinal chemists to identify the fragments responsible for poor drug-like properties and optimize the molecule. The explainability approach was also used to identify the reason behind false positive and false negative MMP predictions. Based on evidence from the existing literature and our analysis, some of these mispredictions were justified. We propose that the quantity, quality, and diversity of the training data will improve the accuracy of property prediction algorithms for novel molecules.

De Novo Structure-Based Drug Design Using Deep Learning.

In recent years, deep learning-based methods have emerged as promising tools for de novo drug design. Most of these methods are ligand-based, where an initial target-specific ligand data set is necessary to design potent molecules with optimized properties. Although there have been attempts to develop alternative ways to design target-specific ligand data sets, availability of such data sets remains a challenge while designing molecules against novel target proteins. In this work, we propose a deep learning-based method, where the knowledge of the active site structure of the target protein is sufficient to design new molecules. First, a graph attention model was used to learn the structure and features of the amino acids in the active site of proteins that are experimentally known to form protein-ligand complexes. Next, the learned active site features were used along with a pretrained generative model for conditional generation of new molecules. A bioactivity prediction model was then used in a reinforcement learning framework to optimize the conditional generative model. We validated our method against two well-studied proteins, Janus kinase 2 (JAK2) and dopamine receptor D2 (DRD2), where we produce molecules similar to the known inhibitors. The graph attention model could identify the probable key active site residues, which influenced the conditional molecule generator to design new molecules with pharmacophoric features similar to the known inhibitors.

An interpretable machine learning model for selectivity of small-molecules against homologous protein family.

Aim: In the early stages of drug discovery, various experimental and computational methods are used to measure the specificity of small molecules against a target protein. The selectivity of small molecules remains a challenge leading to off-target side effects. Methods: We have developed a multitask deep learning model for predicting the selectivity on closely related homologs of the target protein. The model has been tested on the Janus-activated kinase and dopamine receptor families of proteins. Results & conclusion: The feature-based representation (extended connectivity fingerprint 4) with Extreme Gradient Boosting performed better when compared with deep neural network models in most of the evaluation metrics. Both the Extreme Gradient Boosting and deep neural network models outperformed the graph-based models. Furthermore, to decipher the model decision on selectivity, the important fragments associated with each homologous protein were identified.