5 April 2024 in Communications Chemistry
Graph Neural Networks (GNNs) excel in compound property and activity prediction, but the choice of molecular graph representations significantly influences model learning and interpretation. While atom-level molecular graphs resemble natural topology, they overlook key substructures or functional groups and their interpretation partially aligns with chemical intuition. Recent research suggests alternative representations using reduced molecular graphs to integrate higher-level chemical information and leverages both representations for model. However, there is a lack of studies about applicability and impact of different molecular graphs on model learning and interpretation. Here, we introduce MMGX (Multiple Molecular Graph eXplainable discovery), investigating the effects of multiple molecular graphs, including Atom, Pharmacophore, JunctionTree, and FunctionalGroup, on model learning and interpretation with various perspectives. Our findings indicate that multiple graphs relatively improve model performance, but in varying degrees depending on datasets. Interpretation from multiple graphs in different views provides more comprehensive features and potential substructures consistent with background knowledge. These results help to understand model decisions and offer valuable insights for subsequent tasks. The concept of multiple molecular graph representations and diverse interpretation perspectives has broad applicability across tasks, architectures, and explanation techniques, enhancing model learning and interpretation for relevant applications in drug discovery.
23 October 2025 in Communications Chemistry
Designing effective catalysts is a key process for optimizing catalytic reactions to reduce time and waste during scale-up. Recently proposed approaches, including generative models, show promise in identifying new catalysts. However, they are mostly developed for specific reaction classes and predefined fragment categories without considering reaction components, limiting the exploration of novel catalysts across reaction space. Here, we present CatDRX, a catalyst discovery framework powered by a reaction-conditioned variational autoencoder generative model for generating catalysts and predicting their catalytic performance. The model is pre-trained on a broad reaction database and fine-tuned for downstream reactions. Our approach achieves competitive performance in both yield and related catalytic activity prediction. Additionally, it enables effective generation of potential catalysts given reaction conditions by integrating optimization toward desired properties and validation based on reaction mechanisms and chemical knowledge, as demonstrated in various case studies. This work helps facilitate and advance catalyst design and discovery for chemical and pharmaceutical industries.
12 May 2026 in Natural Product Reports
Natural products are an important source of medicines, yet their discovery can be a slow and laborious process.
23 September 2026 in Advanced Intelligent Discovery
Catalyst design has increasingly benefited from advances in AI, particularly chemical language models (CLMs). However, catalysts occupy a distinct chemical space characterized by unique elemental compositions and structural complexity, which limits the applicability of standard CLMs, thereby reducing their utility for broad catalyst discovery. Here, we present CatTransVAE, a catalyst‐specialized CLM based on a transformer variational autoencoder (VAE). CatTransVAE is developed through pretraining on general compounds and subsequent fine‐tuning on diverse catalyst databases. A new template‐guided generation framework is introduced to support controlled catalyst design under structural constraints. CatTransVAE demonstrates promising generative performance across multiple catalyst classes and achieves high task validity in template‐guided generation. In addition, the proposed two‐level embeddings derived from the transformer and VAE components exhibit competitive predictive performance compared with large‐scale CLMs. The practical utility of the framework is demonstrated through a catalyst optimization workflow. These results suggest that CatTransVAE provides a useful catalyst‐specialized representation and generation framework for accelerating catalyst design and discovery.
6 February 2026 in The Journal of Supercomputing
Lead optimization to improve pharmacokinetics and toxicity while maintaining biological activity is an important and costly stage in the drug discovery process, requiring computational approaches for increased efficiency. We propose GraphBioisostere, a bioisostere prediction model that uses graph neural networks. The proposed model leverages a large-scale matched molecular pair dataset constructed from the ChEMBL database and directly learns bioisosterism without target information by considering entire chemical structures. Our evaluation shows that incorporating whole-molecule context improves bioisostere prediction compared to fragment/substituent-only inputs. Compared with a strong fingerprint-based LightGBM baseline, GraphBioisostere achieves competitive prediction performance, with the best GNN variant approaching the baseline ROC-AUC. Additionally, models pre-trained on target-independent bioisostere prediction improved transfer learning performance for potency change prediction against specific targets, particularly in low-data settings. This suggests that GraphBioisostere acquires useful representations of the relationship between chemical structure and activity. Our research provides a tool to evaluate the potential of structural changes in molecular pairs to maintain activity independently of targets, contributing to improved efficiency in the drug discovery process.
17 September 2026 in SLAS Technology