Developing Janus kinase 2 (JAK2) inhibitors has become a significant focus for small-molecule drug discovery programs in recent years because the inhibition of JAK2 may be an effective approach for the treatment of myeloproliferative neoplasm. Here, based on three different types of fingerprints and Extreme Gradient Boosting (XGBoost) methods, we developed three groups of models in that each group contained a classification model and a regression model to accurately acquire highly potent JAK2 kinase inhibitors from the ZINC database. The three classification models resulted in Matthews correlation coefficients of 0.97, 0.94, and 0.97. Docking methods including Glide and AutoDock Vina were employed to evaluate the virtual screening effectiveness of our classification models. The R 2 of three regression models were 0.80, 0.78, and 0.80. Finally, 13 compounds were biologically evaluated, and the results showed that the IC50 values of six compounds were identified to be less than 100 nM. Among them, compound 9 showed high activity and selectivity in that its IC50 value was less than 1 nM against JAK2 while 694 nM against JAK3. The strategy developed may be generally applicable in ligand-based virtual screening campaigns.

Library matching using carbon-13 nuclear magnetic resonance ( 13 C NMR) spectra has been a popular method adopted in compound identification systems. However, the usability of existing approaches has been restricted as enlarging a library containing both a chemical structure and spectrum is a costly and time-consuming process. Therefore, we propose a fundamentally different, novel approach to match 13 C NMR spectra directly against a molecular structure library. We develop a cross-modal retrieval between spectrum and structure (CReSS) system using deep contrastive learning, which allows us to search a molecular structure library using the 13 C NMR spectrum of a compound. In the test of searching 41,494 13 C NMR spectra against a reference structure library containing 10.4 million compounds, CReSS reached a recall@10 accuracy of 91.64% and a processing speed of 0.114 s per query spectrum. When further incorporating a filter with a molecular weight tolerance of 5 Da, CReSS achieved a new remarkable recall@10 of 98.39%. Furthermore, CReSS has potential in detecting scaffolds of novel structures and demonstrates great performance for the task of structural revision. CReSS is built and developed to bridge the gap between 13 C NMR spectra and structures and could be generally applicable in compound identification.

Structure elucidation of unknown compounds based on nuclear magnetic resonance (NMR) remains a challenging problem in both synthetic organic and natural product chemistry. Library matching has been an efficient method to assist structure elucidation. However, it is limited by the coverage of libraries. In addition, prior knowledge such as molecular fragments is neglected. To solve the problem, we propose a conditional molecular generation net (CMGNet) to allow input of multiple sources of information. CMGNet not only uses 13C NMR spectrum data as input but molecular formulas and fragments of molecules are also employed as input conditions. Our model applies large-scale pretraining for molecular understanding and fine-tuning on two NMR spectral data sets of different granularity levels to accommodate structure elucidation tasks. CMGNet generates structures based on 13C NMR data, molecular formula, and fragment information, with a recovery rate of 94.17% in the top 10 recommendations. In addition, the generative model performed well in the generation of various classes of compounds and in the structural revision task. CMGNet has a deep understanding of molecular connectivities from 13C NMR, molecular formula, and fragments, paving the way for a new paradigm of deep learning-assisted inverse problem-solving.

No abstract

In this study, a new sensitivity analysis based on non-linear varying-network magnetic circuit (VNMC) method is proposed for an outer-rotor I-shaped permanent magnet flux-switching motor. By integrating concept of sensitivity analysis into the non-linear VNMC method, the whole design efficiency can be improved effectively by using the comprehensive sensitivity method and sequential non-linear programming algorithm. In the optimisation process, two design objectives are selected, including machine output torque and torque ripple. Based on the sensitivity analysis, the parameters possessing the significant influence on the design objectives can be selected purposely, thus the overall amount of calculation is obviously reduced. After the determination of the sensitive parameters, the rest of optimisation can be realised efficiently by the non-linear VNMC method. Finally, a prototype motor is manufactured and tested. Both theoretical analysis and experimental results confirm the effectiveness of the proposed method.