Springer Science and Business Media LLC

For kinase inhibitors, X-ray crystallography has revealed different types of binding modes. Currently, more than 2000 kinase inhibitors with known binding modes are available, which makes it possible to derive and test machine learning models for the prediction of inhibitors with different binding modes. We have addressed this prediction task to evaluate and compare the information content of distinct molecular representations including protein–ligand interaction fingerprints (IFPs) and compound structure-based structural fingerprints (i.e., atom environment/fragment fingerprints). IFPs were designed to capture binding mode-specific interaction patterns at different resolution levels. Accurate predictions of kinase inhibitor binding modes were achieved with random forests using both representations. The performance of IFPs was consistently superior to atom environment fingerprints, albeit only by less than 10%. An active learning strategy applying information entropy-based selection of training instances was applied as a diagnostic approach to assess the relative information content of distinct representations. IFPs were found to capture more binding mode-relevant information than atom environment fingerprints, leading to highly predictive models even when training instances were randomly selected. By contrast, for atom environment fingerprints, the derivation of accurate models via active learning depended on entropy-based selection of informative training compounds. Notably, higher information content of IFPs confirmed by active learning only resulted in small improvements in global prediction accuracy compared to models derived using atom environment fingerprints. For practical applications, prediction of binding modes of new kinase inhibitors on the basis of chemical structure is highly attractive.

This article presents a quantitative structure–activity relationship (QSAR) approach for predicting the acid dissociation constant (pK $$_a$$ ) of nitrogenous compounds, including those within supramolecular complexes based on cucurbiturils. The model combines low-cost quantum mechanical calculations with QSAR methodology and linear regressions to achieve accurate predictions for a broad range of nitrogen-containing compounds. The model was developed using a diverse dataset of 130 nitrogenous compounds and exhibits excellent predictive performance, with a high coefficient of determination (R $$^2$$ ) of 0.9905, low standard error (s) of 0.3066, and high Fisher statistic (F) of 2142. The model outperforms existing methods, such as Chemaxon software and previous studies, in terms of accuracy and its ability to handle heterogeneous datasets. External validation on pharmaceutical ingredients, dyes, and supramolecular complexes based on cucurbiturils confirms the reliability of the model. To enhance usability, a script-like tool has been developed, providing a streamlined process for users to access the model. This study represents a significant advancement in pK $$_a$$ prediction, offering valuable insights for drug design and supramolecular system optimization.

Idiosyncratic adverse drug reactions have been linked to a drug’s ability to bind with a human leukocyte antigen (HLA) protein. However, due to the thousands of HLA variants and limited structural data for drug-HLA complexes, predicting a specific drug-HLA combination represents a significant challenge. Recently, we investigated the binding mode of abacavir with the HLA-B*57:01 variant using molecular docking. Herein, we developed a new ensemble screening workflow involving three X-ray crystal derived docking procedures to screen the DrugBank database and identify potentially HLA-B*57:01 liable drugs. Then, we compared our workflow’s performance with another model recently developed by Metushi et al., which proposed seven in silico HLA-B*57:01 actives, but were later found to be experimentally inactive. After curation, there were over 6000 approved and experimental drugs remaining in DrugBank for docking using Schrodinger’s GLIDE SP and XP scoring functions. Docking was performed with our new consensus-like ensemble workflow, relying on three different X-ray crystals (3VRI, 3VRJ, and 3UPR) in presence and absence of co-binding peptides. The binding modes of HLA-B*57:01 hit compounds for all three peptides were further explored using 3D interaction fingerprints and hierarchical clustering. The screening resulted in 22 hit compounds forecasted to bind HLA-B*57:01 in all docking conditions (SP and XP with and without peptides P1, P2, and P3). These 22 compounds afforded 2D-Tanimoto similarities being less than 0.6 when compared to the structure of native abacavir, whereas their 3D binding mode similarities varied in a broader range (0.2–0.8). Hierarchical clustering using a Ward Linkage revealed different clustering patterns for each co-binding peptide. When we docked Metushi et al.’s seven proposed hits using our workflow, our screening platform identified six out of seven as being inactive. Molecular dynamic simulations were used to explore the stability of abacavir and acyclovir in complex with peptide P3. This study reports on the extensive docking of the DrugBank database and the 22 HLA-B*57:01 liable candidates we identified. Importantly, comparisons between this study and the one by Metushi et al. highlighted new critical and complementary knowledge for the development of future HLA-specific in silico models.

This publication introduces a novel open-access 31P Nuclear Magnetic Resonance (NMR) shift database. With 14,250 entries encompassing 13,730 distinct molecules from 3,648 references, this database offers a comprehensive repository of organic and inorganic compounds. Emphasizing single-phosphorus atom compounds, the database facilitates data mining and machine learning endeavors, particularly in signal prediction and Computer-Assisted Structure Elucidation (CASE) systems. Additionally, the article compares different models for 31P NMR shift prediction, showcasing the database’s potential utility. Hierarchically Ordered Spherical Environment (HOSE) code-based models and Graph Neural Networks (GNNs) perform exceptionally well with a mean squared error of 11.9 and 11.4 ppm respectively, achieving accuracy comparable to quantum chemical calculations.

The logarithmic acid dissociation constant pKa reflects the ionization of a chemical, which affects lipophilicity, solubility, protein binding, and ability to pass through the plasma membrane. Thus, pKa affects chemical absorption, distribution, metabolism, excretion, and toxicity properties. Multiple proprietary software packages exist for the prediction of pKa, but to the best of our knowledge no free and open-source programs exist for this purpose. Using a freely available data set and three machine learning approaches, we developed open-source models for pKa prediction. The experimental strongest acidic and strongest basic pKa values in water for 7912 chemicals were obtained from DataWarrior, a freely available software package. Chemical structures were curated and standardized for quantitative structure–activity relationship (QSAR) modeling using KNIME, and a subset comprising 79% of the initial set was used for modeling. To evaluate different approaches to modeling, several datasets were constructed based on different processing of chemical structures with acidic and/or basic pKas. Continuous molecular descriptors, binary fingerprints, and fragment counts were generated using PaDEL, and pKa prediction models were created using three machine learning methods, (1) support vector machines (SVM) combined with k-nearest neighbors (kNN), (2) extreme gradient boosting (XGB) and (3) deep neural networks (DNN). The three methods delivered comparable performances on the training and test sets with a root-mean-squared error (RMSE) around 1.5 and a coefficient of determination (R2) around 0.80. Two commercial pKa predictors from ACD/Labs and ChemAxon were used to benchmark the three best models developed in this work, and performance of our models compared favorably to the commercial products. This work provides multiple QSAR models to predict the strongest acidic and strongest basic pKas of chemicals, built using publicly available data, and provided as free and open-source software on GitHub.

Generally, QSAR modelling requires both model selection and validation since there is no a priori knowledge about the optimal QSAR model. Prediction errors (PE) are frequently used to select and to assess the models under study. Reliable estimation of prediction errors is challenging – especially under model uncertainty – and requires independent test objects. These test objects must not be involved in model building nor in model selection. Double cross-validation, sometimes also termed nested cross-validation, offers an attractive possibility to generate test data and to select QSAR models since it uses the data very efficiently. Nevertheless, there is a controversy in the literature with respect to the reliability of double cross-validation under model uncertainty. Moreover, systematic studies investigating the adequate parameterization of double cross-validation are still missing. Here, the cross-validation design in the inner loop and the influence of the test set size in the outer loop is systematically studied for regression models in combination with variable selection. Simulated and real data are analysed with double cross-validation to identify important factors for the resulting model quality. For the simulated data, a bias-variance decomposition is provided. The prediction errors of QSAR/QSPR regression models in combination with variable selection depend to a large degree on the parameterization of double cross-validation. While the parameters for the inner loop of double cross-validation mainly influence bias and variance of the resulting models, the parameters for the outer loop mainly influence the variability of the resulting prediction error estimate. Double cross-validation reliably and unbiasedly estimates prediction errors under model uncertainty for regression models. As compared to a single test set, double cross-validation provided a more realistic picture of model quality and should be preferred over a single test set.