Journal of Cheminformatics最新文献

Molecular representations of chirality, derived from latent space vectors (LSVs) of SMILES heteroencoders, were explored to train machine learning models to predict chiral properties, and were compared to conventional circular fingerprints. Latent space arithmetic was applied to enhance the representation of chirality, by calculating differences between the original descriptor of a molecule and the descriptor of its enantiomer, or the difference between the original descriptor and the descriptor obtained with the stereochemistry-depleted SMILES string. Machine learning was performed with the Random Forest algorithm applied to a dataset of 3858 molecules extracted from the literature (1929 pairs of enantiomers) to predict the elution order observed on the Chiralpak® AD-H column, as well as intrinsic structural chirality labels (R/S or canonical SMILES @/@@). The descriptors derived from the heteroencoders achieved an accuracy of up to 0.75 in the prediction of the elution order, and the fingerprints were superior (0.82). A better predictive ability was observed with the difference LSV descriptors than with the original descriptors.

In this paper, we propose a robust deep-learning model based on a Quantitative Structure − Property Relationship (QSPR) approach for estimating the critical temperature (TC), critical pressure (PC), acentric factor (ACEN) and normal boiling point (NBP) of any C, H, O, N, S, P, F, Cl, Br, I molecule. The Mordred calculator was used to determine 247 descriptors to characterize the molecules considered in this work. For each evaluated property, multiple neural networks were trained within a bagging framework. The predictions from the final ensemble were successfully tested against a large set of experimental data comprising more than 1700 molecules and compared with those from different recent learning models found in the literature. Comprehensive comparisons and extensive testing highlight the robustness and predictive power of the newly proposed multimodal learning model. The developed prediction tool is available on a website at https://lrgp-thermoppt.streamlit.app/. Furthermore, a source code for implementing the trained models in Python is available via github https://github.com/bounac80/AI-ThermPpt.

Natural products provide a rich source of bioactive molecules for a variety of applications. Molecular fingerprints are the tool of choice for systematic large-scale studies of their structures. However, current molecular fingerprints insufficiently represent characteristic features of natural products inherently, decreasing the interpretability of natural product-specific predictions. Here, we show that a natural product-specific molecular fingerprint based on a relatively small set of selected biosynthetic building blocks provides more interpretable predictions of biosynthetic distance and natural product classification. Our fingerprint Biosynfoni outperforms MACCS, Morgan, and Daylight-like fingerprints in biosynthetic distance estimation, using 39 substructure keys. Moreover, Biosynfoni’s design, compactness, and concrete substructure definition allow easy visualisation of the detected substructures and their respective biosynthetic pathway origins. Through Biosynfoni, users can gain more insights from predictions and better examine the importance of features within machine learning models. Our results show that a short fingerprint consisting of biologically significant building blocks performs on par with top-performing molecular fingerprints for natural product classification while improving prediction explainability.

Chemical Language Models (CLMs) have demonstrated capabilities in extracting patterns and predicting from vast volume of the Simplified Molecular Input Line Entry System (SMILES), a notation used to represent molecular structures. Different CLMs, developed from various architectures, can provide unique insights into molecular properties. To harness the uniqueness of different CLMs, we propose FusionCLM, a novel stacking-ensemble learning algorithm that integrate the outputs of multiple CLMs into a unified framework. FusionCLM first generates SMILES embeddings, predictions, and losses from each CLM. Auxiliary models are trained on these first-level predictions and embeddings to estimate test losses during inference. The losses and predictions are then concatenated to create an integrated feature matrix, which trains second-level meta-models for final predictions. Empirical testing on five datasets demonstrates that FusionCLM have better performance than individual CLM at the first level and three advanced multimodal deep learning frameworks, showcasing FusionCLM’s potential in advancing molecular property prediction.

Quantitative structure–activity relationship (QSAR) modeling has become a critical tool in drug design. Recently proposed Topological Regression (TR), a computationally efficient and highly interpretable QSAR model that maps distances in the chemical domain to distances in the activity domain, has shown predictive performance comparable to state-of-the-art deep learning-based models. However, TR’s dependence on simple random sampling-based anchor selection and utilization of radial basis function for response reconstruction constrain its interpretability and predictive capacity. To address these limitations, we propose Adaptive Topological Regression (AdapToR) with adaptive anchor selection and optimization-based reconstruction. We evaluated AdapToR on the NCI60 GI50 dataset, which consists of over 50,000 drug responses across 60 human cancer cell lines, and compared its performance to Transformer CNN, Graph Transformer, TR, and other baseline models. The results demonstrate that AdapToR outperforms competing QSAR models for drug response prediction with significantly lower computational cost and greater interpretability as compared to deep learning-based models.

Retrosynthesis—the process of deconstructing complex molecules into simpler, more accessible precursors—is a cornerstone of drug discovery and material design. While machine learning has improved single-step retrosynthesis prediction, generating complete multi-step retrosynthetic routes remains challenging. In this study, we explore the integration of single-step retrosynthesis models with various planning algorithms to improve multi-step retrosynthetic route generation. We expand the exploration space beyond previously limited settings by incorporating combinations of planning algorithms and single-step retrosynthesis models and diverse datasets, enabling a more comprehensive assessment of retrosynthetic strategies. We evaluated synthetic routes based on both solvability, the ability to generate a complete route, and route feasibility, which reflects their practical executability in the laboratory. Our findings show that the model combination with the highest solvability does not always produce the most feasible routes, underscoring the need for more nuanced evaluation. Through a systematic analysis of combinations of planning algorithms and single-step retrosynthesis models, their performance across different datasets, and various practical metrics, our study provides a more comprehensive evaluation of retrosynthetic planning strategies. These insights contribute to a better understanding of computational retrosynthesis and its alignment with real-world applicability.

When data availability is limited, the prediction of properties through purely data-driven machine learning (ML) is challenging. Integrating physically-based modeling techniques into ML methods may lead to better performance. In a recent work by Chew et al. (“Advancing material property prediction: using physics-informed machine learning models for viscosity”) descriptors from classical molecular dynamics (MD) simulations were included into a quantitative structure–property relationship to accurately predict temperature-dependent viscosity of pure liquids. Through feature importance analysis, the authors found that heat of vaporization was the most relevant descriptor for the prediction of viscosity. In this comment, we would like to discuss the physical origin of this finding by referring to Eyring’s rate theory, and develop an alternative modeling approach using a thermodynamic-based architecture that requires less input data.

Cyclic peptides are promising drug candidates due to their ability to modulate intracellular protein–protein interactions, a property often inaccessible to small molecules. However, their typically poor membrane permeability limits therapeutic applicability. Accurate computational prediction of permeability can accelerate the identification of cell-permeable candidates, reducing reliance on time-consuming and costly experimental screening. Although deep learning has shown potential in predicting molecular properties, its application in permeability prediction remains underexplored. A systematic evaluation of these models is important to assess current capabilities and guide future development. In this study, we conduct a comprehensive benchmark of 13 machine learning models for predicting cyclic peptide membrane permeability. These models cover four types of molecular representations: fingerprints, SMILES strings, molecular graphs, and 2D images. We use experimentally measured PAMPA permeability data from the CycPeptMPDB database, comprising nearly 6000 cyclic peptides, and evaluate performance across three prediction tasks: regression, binary classification, and soft-label classification. Two data-splitting strategies, random split and scaffold split, are used to assess the generalizability of trained models. Our results show that model performance depends strongly on molecular representation and model architecture. Graph-based models, particularly the Directed Message Passing Neural Network (DMPNN), consistently achieve top performance across tasks. Regression generally outperforms classification. Scaffold-based splitting, although intended to more rigorously assess generalization, yields substantially lower model generalizability compared to random splitting. Comparing prediction errors with experimental variability highlights the practical value of current models while also indicating room for further improvement.