Uko Maran

Thyroid hormones (THs) regulate many processes in mammals and, therefore, affect every organ in the body. Thyroid peroxidase (TPO) is an essential enzyme for the successful biosynthesis of THs. Although TPO inhibition is a well-documented molecular initiating event (MIE) in thyroid hormone system disruption adverse outcome pathways (AOPs), experimental methods and computational models to assess TPO activity are lacking. Efficient computational new approach methodologies (NAMs) are a viable solution for identifying TPO inhibitors from a large pool of agrochemicals. The aim of this study was to investigate the suitability of SMILES embeddings generated using a specialized language model (SLM) based on a pretrained deep neural network (DNN) for applying a transfer learning approach in the development of quantitative structure–activity relationships for classifying TPO inhibitors. Traditional theoretical molecular descriptors were used for comparison. Two different molecular descriptor sets resulted in Random Forest (RF) models that performed similarly on the training and test sets, while the sensitivity for the external validation set was substantially different between the two models (0.788 vs 0.490). Comparison of the predictions with the TPO inhibition data of the chemicals assessed by EFSA and EU-NETVAL laboratories showed good agreement. At the same time, analysis of experimental data from other sources showed some conflicting estimates. This suggests that further and more precise studies are needed for some compounds. This study advances in silico methodologies by implementing transfer learning for QSAR modeling from text representations (e.g., SMILES) using the pretrained Bidirectional Encoder Representations from Transformers (BERT) architecture. While traditional QSAR approach relies on molecular descriptors, this evaluation shows that model-generated SMILES embeddings can expand the applicability domain, indicating a more robust representation of structural information compared to traditional molecular descriptors.

Machine Learning (ML) and Artificial Intelligence (AI) approaches have potential to make better-informed decisions in chemical hazard identification while reducing animal testing. Their application in the context of New Approach Methodologies (NAMs) for Hazard Identification in Chemicals Risk Assessment (CRA) is challenging due to the limited knowledge, lack of experience, and uncertainty related to the use of these approaches. Therefore, to facilitate ML and AI approaches' potential acceptance for regulatory use, better standardization, guidelines for transparent reporting, validation, and frameworks are needed to understand their accessibility, verifiability, and usefulness criteria for predictions. An extensive literature review on the availability of ML and AI based NAMs for chemical hazard identification was conducted, focusing primarily on human health endpoints: specific target organ toxicity (STOT), genotoxicity and carcinogenicity, endocrine disruption, skin sensitization, developmental and reproductive toxicity (DART), and repeated dose or chronic toxicity. Nearly 2300 scientific articles were reviewed, and 274 publications with ML-QSAR models revealed that 60.9% of the models described in the scientific literature turned out to be non-usable, 21.9% were potentially usable, and 17.2% were directly usable, i.e., had available software solutions. By endpoint, the skin sensitization is best covered with the ML-QSAR models, followed by endocrine disruption, genotoxicity, and carcinogenicity models. The most derived ML-QSAR models are tree-based models such as random forests, and analogues, followed by artificial neural networks and support vector machine models, with other models being used to a lesser extent. The literature analysis led to a framework that helps model users to identify potentially suitable models for use in a regulatory context. In addition, the framework could help model developers better understand the expectations of model users in a regulatory context and use the framework as a reference when publishing their models, ensuring greater transparency, alignment with regulatory needs, and facilitating future acceptance.

Numerous scientific studies highlight the crucial role of common genetic and epigenetic factors in the development and progression of cancer. To deepen our understanding of how different VEGFR and epigenetic pathways interact in carcinogenesis, the current review examines novel therapeutic agents that target various molecular mechanisms involved in this complex disease. Growing evidence from scientific studies suggests that VEGFR and epigenetic signaling pathways contribute to complex pathophysiological changes in cancer. Therefore, simultaneously targeting VEGFR and epigenetic factors, such as sirtuins, by developing dual inhibitors could provide more individualized therapeutic approaches with safer and more effective outcomes. In this context, Computer-Aided Drug Design (CADD) offers a comprehensive suite of bioinformatic, chemoinformatic, and chemometric approaches to design novel chemotypes of epigenetic dual-target inhibitors. This facilitates the efficient discovery of new drug candidates, enabling innovative treatments for these multifactorial diseases. The review also explores the detailed anticancer mechanisms by which VEGFR, SIRT, and dual-target inhibitors modify metastatic and tumorigenic properties, affect the tumor microenvironment, and regulate the immune response.

Histone deacetylases (HDACs) and Rho-associated coiled-coil containing protein kinases (ROCKs) are critical regulators of tumor development, progression and immunomodulation, particularly in aggressive cancers such as triple negative breast cancer (TNBC). Previous in silico and in vitro studies have demonstrated the therapeutic potential of multitarget HDAC/ROCK inhibition, leading to the identification of C-9 as the first-in-class multitarget HDAC/ROCK inhibitor. In the present study, C-9 served as a lead compound for the rational design of a new series of multitarget inhibitors. Structure-based drug design (SBDD) was used to develop a series of novel HDAC/ROCK multitarget inhibitors. Molecular docking and MM-GBSA binding free energy analysis confirmed strong and specific interactions with the catalytic sites of both targets. Based on these computational insights, ten compounds were synthesized and biologically evaluated using enzyme inhibition and cancer cell assays. Among the synthesized compounds, C-35 and C-40 showed remarkable cytotoxic activity against TNBC cell lines, with IC50 values of 17 µM and 27 µM against MDA-MB-231 cells and about 35 µM against MDA-MB-468 cells, respectively, outperforming known selective HDAC6 and ROCK inhibitors such as tubastatin A and fasudil. Further mechanistic studies revealed that these compounds induce early apoptosis, arrest cell cycle progression and downregulate PD-L1 expression. Remarkably, C-35 also increased the expression of MICA in MDA-MB-231 and MDA-MB-468 cells, possibly promoting the recognition and elimination of tumor cells by immune cells. To further support rational drug optimization, we used synthesized compounds and their IC50 values for the enzymes to form an external blind validation set that enabled the development of machine learning-based quantitative structure-activity relationships (QSAR) models for HDAC6 and ROCK2. These models, developed for the first time for both targets, showed strong predictive performance and were integrated into a comprehensive CADD workflow combining structure-based (molecular docking and MM-GBSA) and ligand-based (QSAR) methods for drug development. Overall, this study highlights the anti-cancer, anti-invasive, anti-migratory and immunomodulatory potential of multitarget HDAC/ROCK inhibition for the treatment of TNBC. Finally, it creates a robust computational-experimental pipeline for future development and optimization of multitarget HDAC/ROCK inhibitors.

The earthworm is a key indicator species in soil ecosystems. This makes the reproductive toxicity of chemical compounds to earthworms a desired property of determination and makes computational models necessary for descriptive and predictive purposes. Thus, the aim was to develop an advanced Quantitative Structure-Activity Relationship modeling approach for this complex property with imbalanced data. The approach integrated gradient-boosted decision trees as classifiers with a genetic algorithm for feature selection and Bayesian optimization for hyperparameter tuning. An additional goal was to analyze and interpret, using SHAP values, the structural features encoded by the molecular descriptors that contribute to pesticide toxicity and nontoxicity, the most notable of which are solvation entropy and a number of hydrolyzable bonds. The final model was constructed as a stacked ensemble of models and combined the strengths of the individual models. Evaluation of this model with an external test set of 147 compounds demonstrated a well-defined applicability domain and sufficient predictive capabilities with a Balanced Accuracy of 77%. The model representation follows FAIR principles and is available on QsarDB.org.

This article presents an in-depth investigation into the influence of anionic structures of ionic liquids (ILs) on gas-ionic liquid partition coefficients (log K) of organic solutes in three ILs. While the primary objective was to examine whether there is a relationship between the molecular structure of the IL anion component and log K, additionally it was looked at whether the molecular descriptors of the anion in the relationships encode possible molecular interactions during the miscibility and partitioning in the IL. The research involves the compilation of data series of experimental log K values, where the cation component is constant. Such representative data series were obtained for three solutes─benzene, cyclohexane, and methanol─in three ILs with a uniform cationic component of methylimidazoliums. Using multiple linear regression models enhanced with machine learning techniques, the relationship between anionic structures and log K values was successfully quantified and modeled. Systematically selected molecular descriptors describing the anion structure show that in the case of methanol log K is strongly dependent on hydrogen bonds and Coulomb-dipolar interactions with the anion component, while in the case of benzene and cyclohexane the dispersion forces of the anion component are dominant. The outlier analysis and data interpretation highlight the need for extensive experimental data. The results confirm the initial hypothesis and provide valuable information on the role of the structure of the anionic component in determining the partitioning behavior of organic solutes. This knowledge is important for the design and optimization of ILs for specific applications, particularly as solvents in various industrial processes. The research also provides useful information about molecular interactions taking place in the interfaces of IL and organic additives in complex liquid media such as multicomponent electrolyte solutions, for example in energy storage applications.

The pyrimidine heterocycle plays an important role in anticancer research. In particular, the pyrimidine derivative families of uracil show promise as structural scaffolds relevant to cervical cancer. This group of chemicals lacks data-driven machine learning quantitative structure-activity relationships (QSARs) that allow for generalization and predictive capabilities in the search for new active compounds. To achieve this, a dataset of pyrimidine and uracil compounds from ChEMBL were collected and curated. A workflow was developed for data-driven machine learning QSAR using an intuitive dataset design and forwards selection of molecular descriptors. The model was thoroughly externally validated against available data. Blind validation was also performed by synthesis and antiproliferative evaluation of new synthesized uracil-based and pyrimidine derivatives. The most active compound among new synthesized derivatives, 2,4,5-trisubstituted pyrimidine was predicted with the QSAR model with differences of 0.02 compared to experimentally tested activity.