Explore the vast range of titles available.

Structure-based virtual screening (SBVS) is a crucial computational approach in drug discovery, but its performance is sensitive to structural variations. Kinases, which are major drug targets, exemplify this challenge due to active site conformational changes caused by different inhibitor types. Most experimentally determined kinase structures have the DFGin state, potentially biasing SBVS towards type I inhibitors and limiting the discovery of diverse scaffolds. We introduce a multi-state modeling (MSM) protocol for AlphaFold2 (AF2) kinase structures using state-specific templates to address these challenges. Our comprehensive benchmarks evaluate predicted model qualities, binding pose prediction accuracy, and hit compound identification through ensemble SBVS. Results demonstrate that MSM models exhibit comparable or improved structural accuracy compared to standard AF2 models, enhancing pose prediction accuracy and effectively capturing kinase-ligand interactions. In virtual screening experiments, our MSM approach consistently outperforms standard AF2 and AF3 modeling, particularly in identifying diverse hit compounds. This study highlights the potential of MSM in broadening kinase inhibitor discovery by facilitating the identification of chemically diverse inhibitors, offering a promising solution to the structural bias problem in kinase-targeted drug discovery.

Predicting the local structural features of a protein from its amino acid sequence helps its function prediction to be revealed and assists in three-dimensional structural modeling. As the sequence-structure gap increases, prediction methods have been developed to bridge this gap. Additionally, as the size of the structural database and computing power increase, the performance of these methods have also significantly improved. Herein, we present a powerful new tool called S-Pred, which can predict eight-state secondary structures (SS8), accessible surface areas (ASAs), and intrinsically disordered regions (IDRs) from a given sequence. For feature prediction, S-Pred uses multiple sequence alignment (MSA) of a query sequence as an input. The MSA input is converted to features by the MSA Transformer, which is a protein language model that uses an attention mechanism. A long short-term memory (LSTM) was employed to produce the final prediction. The performance of S-Pred was evaluated on several test sets, and the program consistently provided accurate predictions. The accuracy of the SS8 prediction was approximately 76%, and the Pearson’s correlation between the experimental and predicted ASAs was 0.84. Additionally, an IDR could be accurately predicted with an F1-score of 0.514. The program is freely available at https://github.com/arontier/S_Pred_Paper and https://ad3.io as a code and a web server.

Structure-based virtual screening (SBVS) is a pivotal computational approach in drug discovery, enabling the identification of potential drug candidates within vast chemical libraries by predicting their interactions with target proteins. The SBVS relies on the receptor protein structures, making it sensitive to structural variations. Kinase, one of the major drug targets, is known as one of the typical examples of an active site conformation change caused by the type of binding inhibitors. Examination of human kinase structures shows that the majority of conformations have the DFGin state. Thus, SBVS using the structures might cause a favor of type of ligand type I inhibitors, bind to the DFGin state, rather than finding the diverse scaffolds. Recent advances in protein structure prediction, such as AlphaFold2 (AF2), offer promising solutions but may still be possibly influenced by the structural bias in existing templates. To address these challenges, we introduce a multi-state modeling (MSM) protocol for kinase structures. We apply MSM to AF2 by providing state-specific templates, allowing us to overcome structural biases and thus apply them to kinase SBVS. We benchmarked our MSM models in three categories: quality of predicted models, reproducibility of ligand binding poses, and identification of hit compounds by ensemble SBVS. The results demonstrate that MSM-generated models exhibit comparable or improved structural accuracy compared to standard AF2 models. We also show that MSM models enhance the accuracy of cognate docking, effectively capturing the interactions between kinases and their ligands. In virtual screening experiments using DUD-E compound libraries, our MSM approach consistently outperforms standard AF2 modeling. Notably, MSM-based ensemble screening excels in identifying diverse hit compounds for kinases with structurally diverse active sites, surpassing standard AF2 models. We highlight the potential of MSM in broadening the scope of kinase inhibitor discovery by facilitating the identification of chemically diverse inhibitors. One of the main problems with structure-based virtual screening is structural flexibility. Ensemble screening is one of the conventional approaches to solving the issue. Gathering experimental structures or molecular simulations could be used to compile the receptor structures. Recent developments in algorithms for predicting protein structures, like AlphaFold2, suggest that different receptor conformations could be produced. However, the prediction approaches produce biased structures because of the bias in the structure database. In order to solve the problem, we developed a protocol called multi-state modeling for kinases. Rather than supplying multiple sequence alignments as an input, we gave the AlphaFold2 a specific template structure and the sequence alignment between the template and query. Our findings imply that our technique can yield a particular structural state of interest with an enhanced or comparable structural quality to AlphaFold2 and predict highly accurate protein-ligand complex structures. Lastly, compared to the typical AlphaFold2 models, ensemble screening using the multi-state modeling approach improves the structure-based virtual screening performance, particularly for diverse active molecular scaffolds.

Respiratory syncytial virus (RSV) remains the leading cause of severe respiratory infections in infants, the elderly, and the immunocompromised. Although stabilized full-length pre-fusion (pre-F) protein vaccines are promising, enhanced respiratory disease (ERD) remains a critical safety concern. Here, we used artificial intelligence to design a de novo immuno-focused antigen that structurally preserves the RSV F head region containing critical neutralising epitopes site zero, II and V while replacing the non-neutralising stem with a computationally designed scaffold to minimise immunopathological risk. The lead candidate, aRF6, elicited robust protective immunity against RSV in mice and similar immunogenicity in non-human primates without detectable toxicity. Importantly, in stringent ERD-promoting models, aRF6 induced minimal pulmonary pathology and markedly attenuated Th2-biased cytokine responses, outperforming formalin-inactivated RSV and full-length-stabilized pre-F. The results of cryoelectron microscopy confirmed that the aRF6 structure precisely matched the computational predictions. These results demonstrated that computationally designed de novo immuno-focused antigens can yield safe and effective RSV vaccines, thereby providing a rational framework for next-generation vaccine development.Competing Interest StatementJ.S., J.L., and J.K. are employee of Arontier Co., Ltd. The authors plan to file a patent application covering aspects of the immunogen design and vaccine strategy reported in this work.