AI Chemistry doi: 10.3390/aichem1040017

Authors: Wenjie Zeng Ran Li Xinyu Wang Junjie Liu Xiaobing Lan Jun Chen

Transition-metal-catalyzed direct C-3 arylation of imidazo[1,2-a]pyridines has witnessed significant advances in recent years. However, identifying optimal reaction conditions for this transformation relies heavily on labor-intensive trial-and-error experiments, which consume significant time and costs. To address this challenge, 275 valid reaction data entries conforming to the reaction mechanism were collected as training data, including substrates, catalysts, additives, solvents, reaction temperature, and experimental yield. Three mainstream tree-based ensemble regression models, including GradientBoosting, Random Forest and ExtraTrees, were constructed after strict data cleaning and advanced feature engineering for systematic benchmark comparison. The GradientBoosting model, which delivered the best fitting and generalization performance among all candidates, was selected as the objective function for subsequent constrained Bayesian optimization. A constrained Bayesian optimization algorithm (with predicted yield bounded within 0–100%) was adopted to globally screen high-dimensional reaction parameters for maximum yield. Comprehensive cross-model evaluation indicated the test-set determination coefficients followed the order GradientBoosting (R2 = 0.4416) > Random Forest (R2 = 0.3864) > ExtraTrees (R2 = 0.3544), while the root mean square error (RMSE) of GradientBoosting was 15.3144, outperforming the other two models in prediction stability. Under the optimal parameter set recommended by GradientBoosting (catalyst Pd(acac)2BuAd2P, reactants ref-5-1d and ref-9-2d, additive CsOAc, solvent DMA, 150 °C), the model predicted a theoretical maximum yield of 100%, whereas Random Forest and ExtraTrees only reached predicted peak yields of 83.63% and 80.80%, respectively. This multi-model benchmarking workflow eliminates the bias risk from single-model prediction, provides an efficient tool for optimizing metal-catalyzed C-3 arylation of imidazo[1,2-a]pyridines, and can be generalized to other complex transition-metal catalytic systems.

AI Chemistry doi: 10.3390/aichem1040016

Authors: Amgad Gerges Una Canning

Somatostatin receptor subtype 2 (SSTR2), encoded at chromosome 17q25.1, is overexpressed in high-risk stage 4 neuroblastoma with a 17q gain, yet no non-peptide small-molecule ligand exists for this indication. From the agonist L-054,522, BROOD generated 5766 analogues, of which 292 were docked to SSTR2 (PDB 7XN9) with FRED, HYBRID, and POSIT; no physicochemical filter was applied. The query achieved an ROC AUC of 0.798 against known ligands and decoys. Eight compounds reaching GREAT pose confidence were re-docked with AutoDock Vina Extended, FITTED, Flare, and DiffDock and then assessed across five ADMET and two retrosynthesis platforms. Redocking the crystallographic ligand recovered the experimental pose with four site-directed programs (RMSD 0.85–1.79 Å), but none run blind or as ensembles, indicating that pose recovery depends on the search protocol rather than the scoring function. Docking scores were not reproducible across repeat runs, so candidates are not ranked. Compound 8 reproduced the Asp122, Gln126, and Thr194 contacts, alone received a Spaya RScore of 1.0 indicating a fully precedented route, and it held its pose over three 100 ns molecular dynamic simulations in a POPC bilayer (ligand RMSD 1.70 ± 0.12 Å). Because docking cannot distinguish agonism from antagonism, this work delivers a synthesisable hypothesis set, not a characterised pharmacological agent.

AI Chemistry doi: 10.3390/aichem1030015

Authors: Viktoria Enkmann Natalija Rajicic

Artificial intelligence is reshaping RNA research by enabling predictive, data-driven workflows across molecular design, therapeutic development, delivery optimization, and analytical quality control. RNA-based medicines, including messenger RNA vaccines, small interfering RNA therapeutics, antisense oligonucleotides, RNA-guided systems, and RNA-targeted small molecules, are programmable but structurally and analytically complex. Their performance depends not only on nucleotide sequence, but also on RNA folding, untranslated regions, chemical modifications, formulation composition, delivery efficiency, stability, manufacturability, and critical quality attributes. This review examines how AI is applied across the RNA therapeutic development pipeline, including RNA sequence and structure prediction, codon optimization, guide RNA design, neoantigen selection, RNA-targeted drug discovery, lipid nanoparticle formulation prediction, and analytical data integration. Particular emphasis is placed on AI-enabled analytics as a translational layer between computational design and therapeutic implementation. Centralized multimodal data platforms that integrate outputs from orthogonal analytical methods can transform fragmented experimental readouts into standardized, model-ready datasets. We argue that future progress will depend on closed-loop systems in which design models, delivery prediction, experimental analytics, and quality-control data continuously inform each other to improve RNA therapeutic translation.

AI Chemistry doi: 10.3390/aichem1030014

Authors: Nagababu Andraju

While Artificial Intelligence (AI) has transformed materials discovery, the primary bottleneck to technological impact remains the transition from lab-scale synthesis to robust, industrial-scale manufacturing. Most promising materials perish in this depth, which is referred as the “valley of death”. The current review consolidates and critically evaluates the emerging ecosystem of AI-driven strategies and frameworks designed specifically to bridge this “lab-to-fab” gap. The review shifts our attention from property prediction to the engineering-driven problems of manufacturability. Furthermore, the review discusses the main obstacles to scaling the production of materials, such as reproducibility, process optimization in the context of uncertainty, and techno-economic viability, as well as the AI approaches being developed to overcome them. This includes Natural Language Processing (NLP) for method extraction, graph neural networks for reaction modeling, reinforcement learning for process control, Bayesian optimization for definition of process windows, and integrated AI–Techno-Economic Analysis (TEA) frameworks. Equally importantly, the review examines the principal failure modes that constrain practical deployment, including out-of-distribution generalization, incomplete and non-transferable literature-derived data, simulator-to-plant mismatch, uncertainty miscalibration, and the continued need for expert oversight. The article concludes with a forward-looking roadmap for the future of AI in chemical and materials engineering. The proposed conclusion is defined by a paradigm shift from simply finding new materials to creating viable, economical, and scalable pathways to produce them, thereby enabling a new era of synthesis-aware materials innovation.

AI Chemistry doi: 10.3390/aichem1030013

Authors: Anzhuo Fan Xiaorong Wang Mingjia Ma Guotao Yang

Multicomponent beef adulteration is challenging to detect rapidly due to high concealment, complex composition, and overlapping spectral features. This study employed visible–near-infrared hyperspectral imaging (400–1000 nm) combined with a Deep Feature-Enhanced Partial Least Squares Regression (DF-PLSR) model for quantitative detection and visualization of adulteration in beef. A total of 2322 samples were prepared across seven adulteration systems—single (chicken, duck, pork), binary, and ternary—with adulteration ratios from 5% to 50%. The DF-PLSR model outperformed traditional PLSR in all systems, achieving the best performance for chicken adulteration (R2p = 0.9942, RMSEP = 1.31%), with RPD > 5 for all systems. CARS and SPA reduced spectral dimensionality by 85.9% while maintaining R2p > 0.93. Pixel-level visualization achieved prediction errors < 3%, enabling intuitive identification of adulterant spatial distribution. This method provides a rapid, non-destructive, and accurate approach for detecting complex adulteration in beef, with strong potential for food safety monitoring.

AI Chemistry doi: 10.3390/aichem1030012

Authors: Didier Mathieu

In the context of artificial intelligence in chemistry, the first applications that come to mind are widely discussed ones, such as building surrogate models to replace costly numerical simulations or designing new compounds. However, one area where generative AI truly stands out is programming assistance, ranging from code improvement suggestions to vibe coding simple programs from scratch. This article explores how these tools can be used by chemists to test new ideas, make the most of existing open-source software and rapidly develop custom solutions that significantly ease day-to-day work. Using the problem of estimating Hansen solubility parameters as an academic example, their strengths and limitations are illustrated through fully customizable sample codes, including a solubility parameter calculator, a molecular spreadsheet library and a molecular editor. In addition to serving as models, these codes could be of practical interest to some readers due to unique features making them address gaps in the current offerings.

AI Chemistry doi: 10.3390/aichem1030011

Authors: Anthony M. Baudino Kari L. Stone

The rapid emergence of metallo-b-lactamase-mediated antibiotic resistance has created an urgent need for new inhibitor discovery strategies. In this work, a machine-learning-guided workflow was developed to generate and prioritize potential inhibitors targeting NDM-1. A SMILES-based variational autoencoder was first pretrained on a broad molecular dataset to learn general chemical syntax and latent molecular representations. The model was then fine-tuned on an 8-hydroxyquinoline-enriched dataset to bias molecular generation toward zinc-binding chemical space relevant to metallo-β-lactamase inhibition. Generated compounds were processed through structural filtering and docking-based evaluation to create training data for downstream predictive modeling. Molecular fingerprints and physicochemical descriptors were then used to train XGBoost models for docking score prediction and classification of potential binders. Classification proved especially useful for prescreening because it avoided overinterpreting small differences in noisy docking scores while still enriching for compounds likely to perform well in docking. The resulting workflow demonstrates how generative modeling and supervised machine learning can be combined to reduce chemical search space, prioritize candidate inhibitors, and guide computational drug discovery. Although experimental validation remains necessary, this approach provides a scalable framework for identifying promising zinc-binding compounds for further molecular simulation and inhibitor development that can be expanded in future studies.

AI Chemistry doi: 10.3390/aichem1030010

Authors: Yunke Yang Ruijie Jiao Siqi Deng Gonghua Hong Junling Guo

Artificial intelligence has greatly accelerated the design and screening of metal–organic materials, particularly for crystalline systems with well-defined topologies and increasingly standardized structural databases. However, this success has also created a structure-centric design paradigm that is less suitable for metal–organic systems whose functions are governed by process history, interfacial assembly, and dynamic coordination rather than by a single idealized lattice. This Perspective proposes that artificial intelligence (AI)-guided design of metal–organic materials should expand beyond crystalline metal–organic frameworks (MOFs) to encompass a broader structural continuum, ranging from long-range ordered frameworks to dynamic, non-periodic coordination networks. Metal–polyphenol networks (MPNs) are used here as an experimentally tractable example within a broader family of structurally dynamic metal–organic materials, as they arise from coordination interactions between metal ions and polyphenolic ligands, generally lack long-range crystallographic periodicity, and exhibit functions that are governed by interfacial assembly, environmental responsiveness, and pathway-dependent structural evolution. These features challenge conventional descriptor design and database-driven prediction, but also create opportunities for AI approaches that are process-aware, interface-sensitive, and function-oriented. By placing MOFs and MPNs within a unified framework of structural order, this Perspective outlines how machine learning, multimodal characterization, active learning, and closed-loop experimentation could expand metal–organic materials design from topology prediction toward dynamic network optimization.

AI Chemistry doi: 10.3390/aichem1020009

Authors: Bocheng Xie Xiaokang Guo Pengwei Xiao Chao Yang

Contrastive learning-based models such as DrugCLIP have recently emerged as scalable tools for structure-based virtual screening by embedding protein structures and small molecules into a shared representation space. While these approaches demonstrate high throughput and competitive screening performance in ligand retrieval tasks, their ability to correctly identify biologically relevant ligand-binding pockets has not been systematically evaluated. Here, we construct a benchmarking dataset comprising 42 pharmacologically diverse human protein targets with experimentally validated drug-bound structures spanning multiple target families. Using this dataset, we evaluate the pocket recognition capability of DrugCLIP and compare its performance with a traditional structure-based workflow (Fpocket combined with ESSA) and a machine learning-based method (P2Rank). DrugCLIP shows robust performance for well-characterized target classes, including kinases (10/10) and nuclear receptors (5/5), but exhibits markedly reduced accuracy for ion channels (1/4), GPCRs (3/5), and transporters (3/5). Notably, pocket prediction accuracy does not strongly correlate with structural data availability, suggesting that intrinsic pocket characteristics rather than training data abundance primarily affect model performance. Across the benchmark, DrugCLIP achieves an overall success rate of 71% (95% CI: 56–83%), compared with 79% (95% CI: 64–88%) for Fpocket+ESSA, and 93% (95% CI: 81–98%) for P2Rank. McNemar’s test showed no significant difference between DrugCLIP and Fpocket+ESSA (p = 0.508), whereas P2Rank significantly outperformed DrugCLIP (p = 0.012). Together, these results provide a quantitative evaluation of pocket recognition by contrastive learning-based models and highlight key limitations of embedding-based approaches for pocket localization.

AI Chemistry doi: 10.3390/aichem1020008

Authors: Davide Zeppilli José Ferraz-Caetano M. Natália D. S. Cordeiro Laura Orian

We designed a supervised machine learning framework to predict standard Gibbs free energies, ΔG°, of formal hydrogen atom transfer (f-HAT) for phenolic antioxidants across different radicals and media, enabling rapid and chemically interpretable screening. We curated a DFT dataset of 71 molecules (phenolic compounds and anthocyanidins), with 207 reaction sites, 10 radical reactive oxygen/sulfur species, and three environments (leading to a total of 6210 ΔG° values). The models amass 106 numerical RDKit descriptors, augmented with one-hot encodings of medium, site, radical, and structural class, and were evaluated through a leave-one-molecule-out protocol. Among the tested regression algorithms, the random forest regressor provides the best balance of accuracy and robustness with both R2 test (≈0.94) and MAE (2.74 kcal mol−1; RMSE (≈5.0 kcal mol−1)), close to DFT chemical accuracy. The feature-importance analysis revealed that “electronic” and “experimental” (site/group) descriptors primarily drive predictions, with the radical’s maximum absolute partial charge being the most important descriptor in the prediction of a radical’s ΔG°. These results suggest that descriptor-driven RF (Random Forest) models can generalize across chemical space to provide interpretable ΔG° predictions, providing a path for chemists towards a scalable route to prioritize antioxidant candidates for broader molecular families.

AI Chemistry doi: 10.3390/aichem1020007

Authors: James Young Dillon Nelson Liping Gu Ruanbao Zhou

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) efficiency constrains carbon fixation, making it a high-value target in biotechnology. The core task of this work is a supervised regression and ranking problem on Rubisco: given a numerical representation of a protein sequence (a PLM embedding), we predict continuous phenotypic scores such as an enzyme kinetic proxy or fitness value. The predictions then guide which variants to test next. Engineering Rubisco is a point of focus but remains challenging due to selection forces in vivo and the combinatorial space of potential mutants for ex vivo uses. We combine protein language model (PLM) embeddings with tabular learning to model Rubisco variant landscapes in two regimes. First, we analyze deep mutational scanning data providing inferred kinetic proxies, including Km for CO2 and Vmax. Second, we model a cyanobacterial screening dataset measuring mutant fitness under differing oxygen and nitrogen regimes, enabling an oxygen tolerance objective. Across tasks, a tabular foundation (TabPFN-2.5) model outperforms gradient-boosted trees on rank-based criteria for variant prioritization, including Spearman correlation and top 5% hit recovery. We then simulate active-learning campaigns initialized with 200 measured variants and iteratively acquiring batches of 48. Model-guided selection recovers more top-performing mutants than random sampling at fixed experimental budgets, even with a conservative XGBoost surrogate. We also demonstrate that Rubisco large-subunit embeddings predict cyanobacterial doubling time and cross-species kinetic parameters, suggesting that Rubisco representation remains meaningful across organisms even with multi-objective cellular constraints. Together, these results support a practical, data-efficient workflow for enzyme engineering and motivate objective-aware design strategies that complement directed evolution.

AI Chemistry doi: 10.3390/aichem1020006

Authors: Shiluo Huang Zheyu Zou

Infrared (IR) spectroscopy is a powerful tool for characterizing molecular structures and chemical groups, offering advantages such as low cost, rapid analysis, and non-destructive testing. When analyzing heterogeneous objects, spectra are typically measured from different regions to capture the local variations, presenting a multi-instance learning (MIL) problem. However, existing methods primarily rely on multi-instance assumptions or explicit bag representations, often failing to fully capture the intrinsic information from implicit representations. We introduce a bag dissimilarity regularized MIL framework for analyzing IR spectra of heterogeneous objects, which integrates both explicit and implicit representations to effectively learn the MIL bags. Specifically, a bag dissimilarity regularization term is utilized to extract implicit representations, which subsequently guide the classifier based on explicit representations to enhance generalization performance. The proposed method was validated on two heterogeneous detection tasks: polydimethylsiloxane (PDMS) block assessment and polyethylene terephthalate (PET) fiber inspection. Experimental results demonstrate that our approach significantly outperforms existing methods on both datasets with a considerable margin.

AI Chemistry doi: 10.3390/aichem1010005

Authors: Matthew A. Oehlschlaeger

Artificial intelligence (AI) is rapidly transforming chemistry and related scientific and engineering disciplines [...]

AI Chemistry doi: 10.3390/aichem1010004

Authors: Benjamin R. Cummins Charles D. Nichols

Modern advances in artificial intelligence have accelerated the development of computational tools for protein–ligand structure prediction, yet their real-world performance remains uneven across receptor classes and ligand chemotypes. Recently published cryo-EM structures of several different psychedelics bound to the serotonin 5HT2A receptor provide a unique opportunity to explore how modern AI-based modeling performs in a pharmacologically important GPCR system. Here, we compare three major approaches: AI-based protein–ligand cofolding (Boltz-2), a leading AI-driven docking module (Uni-Mol Docking v2), and a widely used classical physics-based docking pipeline (AutoDock Vina) across a series of tryptamine and phenethylamine psychedelics. Predicted binding poses were comparatively assessed through structural alignment with these newly available cryo-EM complexes. Additionally, calcium-mobilization assays were performed to provide a coarse functional readout for comparison with computationally predicted binding affinities. This study integrates methodological review with exploratory benchmarking to illustrate how different modeling paradigms behave on a shared receptor–ligand test set. Our results highlight substantial variation between modeling strategies, with AI-based cofolding often producing global binding orientations more closely resembling experimental structures, and classical docking showing greater variability across ligands, while still outperforming AI-driven docking on average. These observations underscore both the growing utility and current limitations of AI-assisted structure prediction in serotonergic drug discovery, and emphasize the importance of careful, experimentally anchored evaluation as such tools continue to advance.

AI Chemistry doi: 10.3390/aichem1010003

Authors: Pengju Yin Xin Li Yuxuan Lv Yan Li Yiping Zhao Bo Hu

Raman spectroscopy has become an important tool for biomedical analysis due to its ability to provide label-free, non-destructive molecular fingerprints of biological samples. However, existing deep learning approaches for classifying biological Raman spectra often focus on specific datasets and lack generalizability and interpretability. In this study, BioRamanNet is presented, an interpretable and generalizable deep learning framework designed for classifying a wide range of biological Raman spectra. The model integrates adaptive one-dimensional convolutional layers and squeeze-and-excitation (SE) blocks within a residual network architecture to enhance feature extraction. BioRamanNet was evaluated using four representative Raman spectral datasets—breast cells, extracellular vesicles and particles (EVPs), viruses, and bacteria—achieving classification accuracies of 99.5%, 100%, 99.8%, and 85.3%, respectively. To improve model interpretability, a perturbation-based analysis using Voigt noise was introduced to identify key wavenumber regions influencing classification. These regions were found to correspond closely with known Raman biomarkers, validating their biological significance. The results of this work demonstrate that BioRamanNet is a powerful and interpretable tool for analyzing diverse biological Raman spectra and holds promise for advancing machine learning-assisted biomedical diagnostics.

AI Chemistry doi: 10.3390/aichem1010002

Authors: Didier Mathieu

Despite the value of molecular packing (MP) calculations in modeling the properties of organic crystals, its widespread adoption is hindered by the absence of a simple tool broadly accessible to non-specialists, and by the lack of reliability inherent to transferable force fields. To fill these gaps, we describe a versatile workflow, leveraging recent progress in the application of machine learning to the parameterization of interatomic potentials. It is provided as a Python script based only on free academic software running on any Linux system. A key ingredient to this workflow is a recent neural network pretrained to predict bespoke force field parameters for any organic compound on the basis of its molecular diagram. The resulting graph-based force field (GB-FF) is fed into the Tinker simulation engine and applied to crystal structures generated using the USPEX crystal structure prediction package. This low-cost workflow is found to outperform current state-of-the-art procedures based on heavily parameterized force fields, thus demonstrating the value of machine-learned bespoke potential parameters.

AI Chemistry doi: 10.3390/aichem1010001

Authors: Olivier Aroule Fabien Torralba Guillaume Hoffmann

Identifying the most reactive conformation of a molecule is a central challenge in computational chemistry, particularly when reactivity depends on subtle conformational effects. While most conformation search tools aim to find the lowest-energy structure, they often overlook the electronic descriptors that govern chemical reactivity. In this work, we present GAELLE, a cheminformatics tool that combines conformer generation with quantum reactivity descriptors to identify the most reactive structure of a molecule in solution. GAELLE integrates an evolutionary algorithm with fast semiempirical quantum chemical calculations (xTB), enabling the automated ranking of conformers based on HOMO–LUMO gap minimization (Pearson’s principle of maximum hardness) and electrophilicity index (Parr’s electrophilicity scale). Solvent effects are accounted for via implicit solvation models (GBSA/ALPB) to ensure realistic evaluation of reactivity in solution. The method is fully SMILES-driven, open-source, and scalable to medium-sized drug-like molecules. Applications to reactive intermediates, bioactive conformations, and pre-reactive complexes demonstrate the method’s relevance for mechanism elucidation, molecular design, and in silico screening. GAELLE is publicly available and offers a reactivity-focused alternative to traditional energy-minimization tools in conformational analysis.