Large language models have greatly enhanced our ability to understand biology and chemistry, yet robust methods for structure-based drug discovery, quantum chemistry and structural biology are still sparse. Precise biomolecule–ligand interaction datasets are urgently needed for large language models. To address this, we present MISATO, a dataset that combines quantum mechanical properties of small molecules and associated molecular dynamics simulations of ~20,000 experimental protein–ligand complexes with extensive validation of experimental data. Starting from the existing experimental structures, semi-empirical quantum mechanics was used to systematically refine these structures. A large collection of molecular dynamics traces of protein–ligand complexes in explicit water is included, accumulating over 170 μs. We give examples of machine learning (ML) baseline models proving an improvement of accuracy by employing our data. An easy entry point for ML experts is provided to enable the next generation of drug discovery artificial intelligence models.

A clinical casein kinase 2 inhibitor, CX-4945 (silmitasertib), shows significant affinity toward the DYRK1A and GSK3β kinases, involved in down syndrome phenotypes, Alzheimer's disease, circadian clock regulation, and diabetes. This off-target activity offers an opportunity for studying the effect of the DYRK1A/GSK3β kinase system in disease biology and possible line extension. Motivated by the dual inhibition of these kinases, we solved and analyzed the crystal structures of DYRK1A and GSK3β with CX-4945. We built a quantum-chemistry-based model to rationalize the compound affinity for CK2α, DYRK1A, and GSK3β kinases. Our calculations identified a key element for CK2α's subnanomolar affinity to CX-4945. The methodology is expandable to other kinase selectivity modeling. We show that the inhibitor limits DYRK1A- and GSK3β-mediated cyclin D1 phosphorylation and reduces kinase-mediated NFAT signaling in the cell. Given the CX-4945's clinical and pharmacological profile, this inhibitory activity makes it an interesting candidate with potential for application in additional disease areas.

Despite decades of efforts, much is still unknown about the hydrolysis of nitrogen dioxide (NO₂), a reaction associated with the formation of acid rain. From the experimental point of view, quantitative analyses are hard, and without pH control the products decompose to some reagents. We resort to high‐level quantum chemistry to compute Gibbs energies for a network of reactions relevant to the hydrolysis of NO₂. With COSMO‐RS solvation corrections we calculate temperature dependent thermodynamic data in liquid water. Using the computed reaction energies, we determine equilibrium concentrations for a gas‐liquid system at controlled pH. For different temperatures and initial concentrations of the different species, we observe that nitrogen dioxide should be fully converted to nitric and nitrous acid. The thermodynamic data in this work can have a potential major impact for several industries with regards to the understanding of atmospheric chemistry and in the reduction of anthropomorphic pollution.

In the pathogenesis of Alzheimer's disease, the overexpression of glycogen synthase kinase-3β (GSK-3β) stands out due to its multifaced nature, as it contributes to the promotion of amyloid β and tau protein accumulation, as well as neuroinflammatory processes. Therefore, in the present study, we have designed, synthesized, and evaluated a new series of GSK-3β inhibitors based on the N-(pyridin-2-yl)cyclopropanecarboxamide scaffold. We identified compound 36, demonstrating an IC₅₀ of 70 nM against GSK-3β. Subsequently, through crystallography studies and quantum mechanical analysis, we elucidated its binding mode and identified the structural features crucial for interactions with the active site of GSK-3β, thereby understanding its inhibitory potency. Compound 36 was effective in the cellular model of hyperphosphorylated tau-induced neurodegeneration, where it restored cell viability after okadaic acid treatment and showed anti-inflammatory activity in the LPS model, significantly reducing NO, IL-6, and TNF-α release. In ADME-tox in vitro studies, we confirmed the beneficial profile of 36, including high permeability in PAMPA (Pe equals 9.4) and high metabolic stability in HLMs as well as lack of significant interactions with isoforms of the CYP enzymes and lack of considerable cytotoxicity on selected cell lines (IC₅₀ > 100 μM on HT-22 cells and 89.3 μM on BV-2 cells). Based on promising pharmacological activities and favorable ADME-tox properties, compound 36 may be considered a promising candidate for in vivo research as well as constitute a reliable starting point for further studies.

Large language models (LLMs) have greatly enhanced our ability to understand biology and chemistry. Yet, relatively few robust methods have been reported for structure-based drug discovery. Highly precise biomolecule-ligand interaction datasets are urgently needed in particular for LLMs, that require extensive training data. We present MISATO, the first dataset that combines quantum mechanics properties of small molecules and associated molecular dynamics simulations of about 20000 experimental protein-ligand complexes. Starting from the PDBbind dataset, semi-empirical quantum mechanics was used to systematically refine these structures. The largest collection to date of molecular dynamics traces of protein-ligand complexes in explicit water are included, accumulating to 170 μs. We give ML baseline models and simple Python data loaders, and aim to foster a thriving community around MISATO ( https://github.com/t7morgen/misato-dataset ). An easy entry point for ML experts is provided without the need of deep domain expertise to enable the next generation of drug discovery AI models.