{"title":"Predicting adverse drug reactions for combination pharmacotherapy with cross-scale associative learning via attention modules.","authors":"Boyang Li, Yifan Qi, Bo Li, Xiaoqiong Li","doi":"10.1038/s43588-025-00816-7","DOIUrl":"https://doi.org/10.1038/s43588-025-00816-7","url":null,"abstract":"<p><p>The rapid emergence of combination pharmacotherapies offers substantial therapeutic advantages but also poses risks of adverse drug reactions (ADRs). The accurate prediction of ADRs with interpretable computational methods is crucial for clinical medication management, drug development and precision medicine. Machine-learning and recently developed deep learning architectures struggle to effectively elucidate the key protein-protein interactions underlying ADRs from an organ perspective and to explicitly represent ADR associations. Here we propose OrganADR, an associative learning-enhanced model to predict ADRs at the organ level for emerging combination pharmacotherapy. It incorporates ADR information at the organ level, drug information at the molecular level and network-based biomedical knowledge into integrated representations with multi-interpretable modules. Evaluation across 15 organs demonstrates that OrganADR not only achieves state-of-the-art performance but also delivers both interpretable insights at the organ level and network-based perspectives. Overall, OrganADR represents a useful tool for cross-scale biomedical information integration and could be used to prevent ADRs during clinical precision medicine.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144531465","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Alberto Megías, Sergio Contreras Arredondo, Cheng Giuseppe Chen, Chenyu Tang, Benoît Roux, Christophe Chipot
{"title":"Iterative variational learning of committor-consistent transition pathways using artificial neural networks.","authors":"Alberto Megías, Sergio Contreras Arredondo, Cheng Giuseppe Chen, Chenyu Tang, Benoît Roux, Christophe Chipot","doi":"10.1038/s43588-025-00828-3","DOIUrl":"https://doi.org/10.1038/s43588-025-00828-3","url":null,"abstract":"<p><p>Discovering transition pathways that are physically meaningful and committor-consistent has long been a challenge in studying rare events in complex systems. Here we introduce a neural network-based strategy that learns simultaneously the committor function and the associated committor-consistent string, offering an unprecedented view of transition processes. Built on the committor time-correlation function, this method operates across diverse dynamical regimes, and extends beyond traditional approaches relying on infinitesimal time-lag approximations, valid only in the overdamped diffusive limit. It also distinguishes multiple competing pathways, crucial for understanding complex biomolecular transformations. Demonstrated on benchmark potentials and biological systems such as peptide isomerization and protein-model folding, this approach robustly reproduces established dynamics, rate constants and transition mechanisms. Its adaptability to collective variables and resilience across neural architectures make it a powerful and versatile tool for enhanced-sampling simulations of rare events, enabling insights into the intricate landscapes of biomolecular systems.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144531464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yang Zhou, Qiongyu Sheng, Guohua Wang, Li Xu, Shuilin Jin
{"title":"Quantifying batch effects for individual genes in single-cell data.","authors":"Yang Zhou, Qiongyu Sheng, Guohua Wang, Li Xu, Shuilin Jin","doi":"10.1038/s43588-025-00824-7","DOIUrl":"10.1038/s43588-025-00824-7","url":null,"abstract":"<p><p>Batch effects substantially impede the comparison of multiple single-cell experiment batches. Existing methods for batch effect removal and quantification primarily emphasize cell alignment across batches, often overlooking gene-level batch effects. Here we introduce group technical effects (GTE)-a quantitative metric to assess batch effects on individual genes. Using GTE, we show that batch effects unevenly impact genes within the dataset. A portion of highly batch-sensitive genes (HBGs) differ between datasets and dominate the batch effects, whereas non-HBGs exhibit low batch effects. We demonstrate that as few as three HBGs are sufficient to introduce substantial batch effects. Our method also enables the assessment of cell-level batch effects, outperforming existing batch effect quantification methods. We also observe that biologically similar cell types undergo similar batch effects, informing the development of data integration strategies. The GTE method is versatile and applicable to various single-cell omics data types.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144512906","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"The impact of language models on the humanities and vice versa.","authors":"Ted Underwood","doi":"10.1038/s43588-025-00819-4","DOIUrl":"https://doi.org/10.1038/s43588-025-00819-4","url":null,"abstract":"","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499776","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hongli Ma, Letian Gao, Yunfan Jin, Jianwei Ma, Yilan Bai, Xiaofan Liu, Pengfei Bao, Ke Liu, Zhenjiang Zech Xu, Zhi John Lu
{"title":"RNA-ligand interaction scoring via data perturbation and augmentation modeling.","authors":"Hongli Ma, Letian Gao, Yunfan Jin, Jianwei Ma, Yilan Bai, Xiaofan Liu, Pengfei Bao, Ke Liu, Zhenjiang Zech Xu, Zhi John Lu","doi":"10.1038/s43588-025-00820-x","DOIUrl":"https://doi.org/10.1038/s43588-025-00820-x","url":null,"abstract":"<p><p>Despite recent advances in RNA-targeting drug discovery, the development of data-driven deep learning models remains challenging owing to limited validated RNA-small molecule interaction data and scarce known RNA structures. In this context, we introduce RNAsmol, a sequence-based deep learning framework that incorporates data perturbation with augmentation, graph-based molecular feature representation and attention-based feature fusion modules to predict RNA-small molecule interactions. RNAsmol employs perturbation strategies to balance the bias between the true negative and unknown interaction space, thereby elucidating the intrinsic binding patterns between RNA and small molecules. The resulting model demonstrates accurate predictions of the binding between RNA and small molecules, outperforming other methods in ten-fold cross-validation, unseen evaluation and decoy evaluation. Moreover, we use case studies to visualize molecular binding profiles and the distribution of learned weights, providing interpretable insights into RNAsmol's predictions. In particular, without requiring structural input, RNAsmol can generate reliable predictions and be adapted to various drug design scenarios.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144487462","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Robert MacKnight, Daniil A Boiko, Jose Emilio Regio, Liliana C Gallegos, Théo A Neukomm, Gabe Gomes
{"title":"Rethinking chemical research in the age of large language models.","authors":"Robert MacKnight, Daniil A Boiko, Jose Emilio Regio, Liliana C Gallegos, Théo A Neukomm, Gabe Gomes","doi":"10.1038/s43588-025-00811-y","DOIUrl":"https://doi.org/10.1038/s43588-025-00811-y","url":null,"abstract":"<p><p>Large language models (LLMs) offer opportunities for advancing chemical research, including planning, optimization, data analysis, automation and knowledge management. Deploying LLMs in active environments, where they interact with tools and data, can greatly enhance their capabilities. However, challenges remain in evaluating their performance and addressing ethical issues such as reproducibility, data privacy and bias. Here we discuss ongoing and potential integrations of LLMs in chemical research, highlighting existing challenges to guide the effective use of LLMs as active scientific partners.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144487461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Cost-effective instruction learning for pathology vision and language analysis.","authors":"Kaitao Chen, Mianxin Liu, Fang Yan, Lei Ma, Xiaoming Shi, Lilong Wang, Xiaosong Wang, Lifeng Zhu, Zhe Wang, Mu Zhou, Shaoting Zhang","doi":"10.1038/s43588-025-00818-5","DOIUrl":"10.1038/s43588-025-00818-5","url":null,"abstract":"<p><p>The advent of vision-language models fosters interactive conversations between artificial intelligence-enabled models and humans. However, applying these models in the clinic faces challenges related to large-scale training data as well as financial and computational resources. Here we propose CLOVER, a cost-effective instruction learning framework for conversational pathology. CLOVER trains a lightweight module and uses instruction tuning while freezing the parameters of the large language model. Instead of using costly GPT-4, we propose well-designed prompts on GPT-3.5 for building generation-based instructions, emphasizing the utility of pathological knowledge derived from the Internet source. We construct a high-quality set of template-based instructions in the context of digital pathology. Using two benchmark datasets, our findings reveal the strength of hybrid-form, pathological visual question-answer instructions. CLOVER outperforms baselines that possess 37 times more training parameters and exhibits few-shot capacity on an external clinical dataset. CLOVER could thus accelerate the adoption of rapid conversational applications in digital pathology.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144334587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hengwei Bian, Xueguang Shao, Christophe Chipot, Wensheng Cai, Haohao Fu
{"title":"A formally exact method for high-throughput absolute binding-free-energy calculations.","authors":"Hengwei Bian, Xueguang Shao, Christophe Chipot, Wensheng Cai, Haohao Fu","doi":"10.1038/s43588-025-00821-w","DOIUrl":"https://doi.org/10.1038/s43588-025-00821-w","url":null,"abstract":"<p><p>Here we introduce a high-throughput, formally exact method for absolute binding-free-energy calculations that enhances computational efficiency and accuracy. At the core of this method is a thermodynamic cycle that minimizes protein ligand relative motion, thereby reducing system perturbations and driving a fourfold gain in efficiency over the traditional double-decoupling method. By combining this strategy with double-wide sampling and hydrogen-mass repartitioning algorithms, the efficiency is further boosted to eightfold. The presented method is applied to 45 diverse protein-ligand complexes. For 34 complexes with validated force-field accuracy, our method achieves an average unsigned error of less than 1 kcal mol<sup>-1</sup> and a hysteresis below 0.5 kcal mol<sup>-1</sup>, showcasing exceptional reliability. Moreover, it efficiently manages flexible peptide ligands through a potential-of-mean-force calculation, adding less than 5% extra simulation time. For 11 challenging cases, the presented method also shows an improvement compared with previously published results. Put together, this method has potential for advancing research in physical, biological and medicinal chemistry.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144328015","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Bojun Liu, Siqin Cao, Jordan G Boysen, Mingyi Xue, Xuhui Huang
{"title":"Memory kernel minimization-based neural networks for discovering slow collective variables of biomolecular dynamics.","authors":"Bojun Liu, Siqin Cao, Jordan G Boysen, Mingyi Xue, Xuhui Huang","doi":"10.1038/s43588-025-00815-8","DOIUrl":"https://doi.org/10.1038/s43588-025-00815-8","url":null,"abstract":"<p><p>Identifying collective variables (CVs) that accurately capture the slowest timescales of protein conformational changes is crucial to comprehend numerous biological processes. Here we introduce memory kernel minimization-based neural networks (MEMnets), a deep learning framework that accurately identifies the slow CVs of biomolecular dynamics. Unlike popular CV-identification methods, which typically assume Markovian dynamics, MEMnets is built on the integrative generalized master equation theory, which incorporates non-Markovian dynamics by encoding them in a memory kernel for continuous CVs. The key innovation of MEMnets is the identification of optimal CVs by minimizing the upper bound for the time-integrated memory kernels through parallel encoder networks. We demonstrate that MEMnets effectively identifies slow CVs involved in the folding of the FIP35 WW domain, revealing two parallel folding pathways. In addition, we illustrate MEMnets' robust numerical stability in identifying meaningful CVs in large biomolecular dynamic systems with limited sampling by applying it to the clamp opening of bacterial RNA polymerase, a much more complex conformational change.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144268053","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}