Química Inorgànica i Orgànica

Transfer learning based on atomic feature extraction for the prediction of experimental ¹³C chemical shifts

2026-03-03T02:34:36Z

Transfer learning based on atomic feature extraction for the prediction of experimental ¹³C chemical shifts Ivković, Žarko; Jover Modrego, Jesús; Harvey, Jeremy Forecasting experimental chemical shifts of organic compounds is a long-standing challenge in organic chemistry. Recent advances in machine learning (ML) have led to routines that surpass the accuracy of ab initio Density Functional Theory (DFT) in estimating experimental 13C shifts. The extraction of knowledge from other models, known as transfer learning, has demonstrated remarkable improvements, particularly in scenarios with limited data availability. However, the extent to which transfer learning improves predictive accuracy in low-data regimes for experimental chemical shift predictions remains unexplored. This study indicates that atomic features derived from a message passing neural network (MPNN) forcefield are robust descriptors for atomic properties. A dense network utilizing these descriptors to predict 13C shifts achieves a mean absolute error (MAE) of 1.68 ppm. When these features are used as node labels in a simple graph neural network (GNN), the model attains a better MAE of 1.34 ppm. On the other hand, embeddings from a self-supervised pre-trained 3D aware transformer are not sufficiently descriptive for a feedforward model but show reasonable accuracy within the GNN framework, achieving an MAE of 1.51 ppm. Under low-data conditions, all transfer-learned models show a significant improvement in predictive accuracy compared to existing literature models, regardless of the sampling strategy used to select from the pool of unlabeled examples. We demonstrated that extracting atomic features from models trained on large and diverse datasets is an effective transfer learning strategy for predicting NMR chemical shifts, achieving results on par with existing literature models. This method provides several benefits, such as reduced training times, simpler models with fewer trainable parameters, and strong performance in low-data scenarios, without the need for costly ab initio data of the target property. This technique can be applied to other chemical tasks opening many new potential applications where the amount of data is a limiting factor.

Colistin-loaded biodegradable nanoparticles as a promising antibacterial medication to reduce colistin-induced toxicity

2026-02-20T20:01:53Z

Colistin-loaded biodegradable nanoparticles as a promising antibacterial medication to reduce colistin-induced toxicity Cano Fernández, Amanda; Ettcheto Arriola, Miren; Sánchez-López, E. (Elena); Guzman, Laura; Segovia, Roser; Espina García, Marta; Carbó Banús, Marcel·lí; Olloquequi, Jordi; Barenys Espadaler, Marta; Cajal Visa, Yolanda; Camins, Àngels; García López, María Luisa; Rabanal Anglada, Francesc Infectious diseases cause mortality rates over 17 million people per year. Among them, bacterial infections constitute one of the major causes. Pneumonia and nosocomial infections are the most severe bacterial infections. Moreover, the indiscriminate use of antibiotics during the last decades has triggered an increasing multiple resistance towards these drugs, which represents a serious global socioeconomic and public health risk. In this sense, nanomedicine has provided an innovative therapeutic alternative able to accumulate the drug in the site of the infection, improve its effectiveness and reduce the inherent toxicity, thus leading to overcoming bacterial resistance. In this work, we aimed to encapsulate colistin, an antibiotic commonly used against multi-drug resistant bacteria in polymeric nanoparticles of poly(lactic-co-glycolic) acid (COL-NPs). COL-NPs were optimized obtaining an average size below 200 nm, monodisperse population and a negative surface charge. Physicochemical assays confirmed that the drug was encapsulated into the polymeric matrix and COL-NPs possessed a round shape and a smooth surface. Moreover, COL-NPs were able to release the drug in a sustained manner and showed suitable stability. In addition, in vitro assays confirmed that COL-NPs were effective against different gram-negative bacterial species such as Pseudomonas aeruginosa, Escherichia coli and Acinetobacter baumannii. Finally, in vivo experiments showed that COL-NPs did not promote any toxicological effects in treated mice, reducing renal concentrations compared to the free drug, maintaining urea levels comparable to those of the control group, and decreasing most of the colistin-induced neurotoxic effects. All these results together suggest that COL-NPs could be a promising therapeutic tool against drug-resistant bacterial infections.

Rhodium(I) Complexes with a n1- Fluorenyl, P-Phosphanyl-Phosphorane Ligand

2026-02-19T20:10:07Z

Rhodium(I) Complexes with a n1- Fluorenyl, P-Phosphanyl-Phosphorane Ligand Eusamio, Javier; Saumell Palacios, Nil; Vidal Ferran, Anton; Grabulosa, Arnald The first example of a P-phosphanylphosphorane, Flu�PCy2−PCy2 (L2; Flu = 9-fluorenyl), has been easily prepared by P-phosphination of lithiated 9-dicyclohexylphosphinofluorene (FluPCy2, L0) with chlorodicyclohexylphosphane. L2 constitutes a new type of P(III)−P(V) organophosphorus compound, a σ3 λ3 −σ4 λ5 species that is stable under an inert atmosphere in the solid state. The reaction of L2 with [Rh(diene)2]BR4 causes metalation of the benzylic carbon (C9) of fluorene, giving κ2 -C,P complexes in which fluorene is coordinated in the η1 form. A complex with the weakly coordinating BArF anion has been isolated and fully characterized, including its crystal structure obtained by X-ray diffraction.

Self-assembly of a supramolecular spin-crossover tetrahedron

2026-02-03T19:32:13Z

Self-assembly of a supramolecular spin-crossover tetrahedron Nielsen, Hannah H.; Vilarino Casaus, Pol; Rodriguez, Gemma; Trepard, Florian; Roubeau, Olivier; Aromí Bedmar, Guillem; Aguilà Avilés, David Spin-crossover (SCO) compounds are fascinating switchable materials with great potential for the development of novel technological devices. These coordination complexes exhibit metal ions with two possible electronic configurations (low- spin, LS, and high-spin, HS) which can be toggled using exter- nal stimuli such as temperature, pressure, or light irradiation. The different magnetic, optical, and structural features of the two states allow these materials to be exploited for a wide range of applications, such as sensors, actuators, or for information storage. Interestingly, the physical pro- perties of SCO compounds can be tuned by modifying the weak non-covalent interactions exhibited within or in between their molecular entities. In host–guest systems, these inter- actions offer a versatile tool, for example, for manipulating the transition temperature of encapsulating SCO complexes simply by altering the nature of the supramolecular guest, as shown in dinuclear helicates, tetrahedral cages, or cubic architec- tures. Long range intermolecular interactions can be exploited as well to tune or even to activate/deactivate the SCO behaviour. Such modulation arises from the nature and strength of such interaction, which influence the communi- cation between molecules and thus its cooperativity, or affect the ligand field exerted by the donor set and therefore the SCO temperature.

Química Inorgànica i Orgànica

Transfer learning based on atomic feature extraction for the prediction of experimental 13C chemical shifts

Colistin-loaded biodegradable nanoparticles as a promising antibacterial medication to reduce colistin-induced toxicity

Rhodium(I) Complexes with a n1- Fluorenyl, P-Phosphanyl-Phosphorane Ligand

Self-assembly of a supramolecular spin-crossover tetrahedron

Transfer learning based on atomic feature extraction for the prediction of experimental ¹³C chemical shifts