Language：English   Publisher：Communications Chemistry  

The scarcity of experimental training data restricts the integration of machine learning into catalysis research. Here, we report on the effectiveness of graph convolutional network (GCN) models pretrained on a molecular topological index, which is not used in typical organic synthesis, for estimating the catalytic activity, a task that usually requires high levels of human expertise. For pretraining, we used custom-tailored virtual molecular databases that can be readily constructed using either a systematic generation method or a molecular generator developed in our group. Although 94%–99% of the employed virtual molecules are unregistered in the PubChem database, the resulting pretrained GCN models improve the prediction of catalytic activity for real-world organic photosensitizers. The results demonstrate the efficiency of the present transfer-learning strategy, which leverages readily obtainable information from self-generated virtual molecules. (Figure presented.)

DOI： 10.1038/s42004-025-01678-w

Web of Science

Scopus

PubMed

Language：English   Publisher：Nature Communications  

While seasoned organic chemists can often predict suitable catalysts for new reactions based on their past experiences in other catalytic reactions, developing this ability is costly, laborious and time-consuming. Therefore, replicating this remarkable expertize of human researchers through machine learning (ML) is compelling, albeit that it remains highly challenging. Herein, we apply a domain-adaptation-based transfer-learning (TL) approach to photocatalysis. Despite being different reaction types, the knowledge of the catalytic behavior of organic photosensitizers (OPSs) from photocatalytic cross-coupling reactions is successfully transferred to ML for a [2+2] cycloaddition reaction, improving the prediction of the photocatalytic activity compared with conventional ML approaches. Furthermore, a satisfactory predictive performance is achieved by using only ten training data points. This experimentally readily accessible small dataset can also be used to identify effective OPSs for alkene photoisomerization, thereby showcasing the potential benefits of TL in catalyst exploration.

DOI： 10.1038/s41467-025-58687-5

Open Access

Web of Science

Scopus

PubMed

Language：English   Publisher：Chemistry A European Journal  

Herein, we report that the catalytic activity of phosphorus compounds with a P─H bond in dehydrative amidation reactions is substantially enhanced when using catechol derivatives as additives. A systematic investigation into the catalytic activity of four phosphorus compounds and 17 catechol derivatives allowed identifying several effective combinations, for example, dimethyl phosphite and 2,3-dihydroxynaphthalene, for the catalytic amidation of carboxylic acids with amines. The developed catalytic system enables the synthesis of pharmaceuticals and facilitates the amidation of anilines, which are typically considered poor substrates with low reactivity. In addition, machine-learning (ML) models were constructed to validate the effectiveness of the database generated from the above-mentioned investigation. Multilayer perceptron (MLP) models incorporating descriptors derived from both quantum chemical calculations and RDKit provided accurate predictions while maintaining chemical interpretability, which was underscored by a visual analysis using SHAP.

DOI： 10.1002/chem.202500955

Web of Science

Scopus

PubMed

Language：English   Publisher：Journal of Organic Chemistry  

During the recent development of machine-learning (ML) methods for organic synthesis, the value of “failed experiments” has increasingly been acknowledged. Accordingly, we have developed an exhaustive database comprising 300 entries of experimental data obtained by performing ruthenium-catalyzed hydrogenation reactions using 10 ketones as substrates and 30 phosphine ligands. After evaluating the predictive performance of ML models using the constructed database, we conducted a virtual screening of commercially available phosphine ligands. For the virtual screening, we utilized several models, such as histogram-based gradient boosting and Ridge regression, combined with the Mordred descriptors and MACCSKeys, respectively. The disclosed approach resulted in the identification of high-performance phosphine ligands, and the rationale behind the predictions in the virtual screening was analyzed using SHAP.

DOI： 10.1021/acs.joc.4c02347

Web of Science

Scopus

PubMed

Publisher：Organometallics  

Herein, we demonstrate the significant impact of tetradentate-ligand-coordinated metal complexes, which have not yet been exploited for the direct catalytic hydrogenation of carboxylic acids (CAs). Our previously developed cationic iridium complex coordinated with a PNNP-ligand [(PNNP)Ir] is effective for hydrogenating esters and carboxylic anhydrides generated in situ from CAs but unsuitable for the direct hydrogenation of CAs. In sharp contrast, the corresponding neutral iridium complex with a PNCP-ligand [(PNCP)Ir] developed in this study facilitates the direct hydrogenation of CAs, including biorelevant and pharmaceutical compounds, under not more than 1 MPa of H<inf>2</inf>. Quantum-chemical calculations indicated that (PNCP)Ir is kinetically a far more competent catalyst than (PNNP)Ir, particularly for the C-H bond formation via hydride transfer from Ir-H to the carbonyl carbon of CA, which was identified as the rate-determining step. While Ir-carboxylates are in resting states throughout the catalytic cycle, CA itself barely interacts with the Ir center during the hydride transfer process.

DOI： 10.1021/acs.organomet.4c00355

Web of Science

Scopus