The rise of deep learning in various scientific and technology areas promotes the development of AI‐based tools for information retrieval. Optical recognition of organic structures is a key part of the automated extraction of chemical information. However, this is a challenging task because there is a large variety of representation styles. In this research, we present a Transformer‐based artificial neural network to convert images of organic structures to molecular structures. To train the model, we created a comprehensive data generator that stochastically simulates various drawing styles, functional groups, functional group placeholders (R‐groups), and visual contamination. We demonstrate that the Transformer‐based architecture can gather chemical insights from our generator with almost absolute confidence. That means that, with Transformer, one can fully concentrate on data simulation to build a good recognition model. A web demo of our optical recognition engine is available online at Syntelly platform, and the code for dataset generation is available on GitHub.

Solubility is one of the key properties of organic compounds that determines their applications in chemistry, materials science and pharmaceuticals. However, predicting solubility values in any solvent except water from a molecular structure still remains a challenging task in modern cheminformatics, not least due to the lack of large and diverse datasets. In this study, we present a dataset containing 103944 experimental solubility values within a temperature range from 243 to 425 K for 1448 organic compounds measured in 213 individual solvents extracted from 1595 peer-reviewed articles. The molecular structures of solutes and solvents as well as solubility data are standardized and provided in a machine-readable format, allowing straightforward data-driven analysis. We have also developed a web-tool for interactive visualization and search within the dataset. This dataset can serve as a comprehensive benchmark for developing machine learning for predicting solubility.

Iridium(III) complexes nowadays became rising stars in various health-related applications. Thus, there is a necessity to assess cytotoxicity of the synthesized molecules against cancer/normal cell lines. In this report, we present a dataset of 2694 experimental cytotoxicity values of 803 iridium complexes against 127 different cell lines. We specify the experimental conditions and provide representation of the complexes molecules in machine-readable format. The dataset provides a starting point for exploration of new iridium-based cellular probes and opens new possibilities for predictions of toxicities and data-driven generation of new organometallic anticancer agents.

Solubility is crucial in organic chemistry and holds significant value in the field of medicinal chemistry. Employing computational and QSPR modeling for solubility estimation is favorable as it reduces experimental costs. However, high-quality experimental data is essential for training these QSPR models. In our study, we compiled a dataset consisting of 54,273 experimental solubility values within a temperature range of 243.15 to 403.15 K in various organic solvents and water. This dataset can be used as a reference for individual values or training solubility QSPR models. We conducted a statistical analysis and identified prevalent patterns in the data. Furthermore, we developed an interactive, parametric t-SNE-based tool to explore the chemical space of solutes. Utilizing this tool, we characterized common scaffolds in the dataset and demonstrated that the chemical space of solutes is extensive and diverse.

Cyclometalated iridium(III) complexes are excellent emitters for organic light-emitting diodes (OLEDs), but the design of these compounds requires substantial cost and experimental efforts. In this work, we aimed at a simple and fast data-driven prediction of their luminescence properties based on the experimental data. To this end, we created a database (IrLumDB) that contains experimentally measured luminescence properties for over 1200 structurally diverse bis-cyclometalated iridium(III) complexes from 340 peer-reviewed articles. Based on the IrLumDB, we trained machine learning models, which require only simplified molecular input line entry system (SMILES) of ligands, to accurately predict the wavelength of emission maxima (λ_max) and photoluminescence quantum yield (PLQY) for the iridium phosphors. We also developed a PLQY classifier, which allows identification of highly efficient emitters (PLQY > 0.5). Furthermore, we validated the models for λ_max prediction on the set of 33 experimentally obtained luminescence spectra for newly synthesized and characterized Ir(III) complexes. Our methodology will serve as a straightforward approach for the prediction of the luminescence properties of large sets of bis-cyclometalated iridium(III) complexes, enabling identification of potentially efficient phosphors for subsequent synthesis. The properties prediction and the IrLumDB exploration are available at https://irlumdb.streamlit.app/.