Pub Date : 2025-09-26DOI: 10.1038/s43588-025-00878-7
Zhe Liu (, ), Yihang Bao (, ), An Gu (, ), Weichen Song (, ), Guan Ning Lin (, )
Noncoding mutations play a critical role in regulating gene expression, yet predicting their effects across diverse tissues and cell types remains a challenge. Here we present EMO, a transformer-based model that integrates DNA sequence with chromatin accessibility data (assay for transposase-accessible chromatin with sequencing) to predict the regulatory impact of noncoding single nucleotide polymorphisms on gene expression. A key component of EMO is its ability to incorporate personalized functional genomic profiles, enabling individual-level and disease-contextual predictions and addressing critical limitations of current approaches. EMO generalizes across tissues and cell types by modeling both short- and long-range regulatory interactions and capturing dynamic gene expression changes associated with disease progression. In benchmark evaluations, the pretraining-based EMO framework outperformed existing models, with fine-tuning small-sample tissues enhancing the model’s ability to fit target tissues. In single-cell contexts, EMO accurately identified cell-type-specific regulatory patterns and successfully captured the effects of disease-associated single nucleotide polymorphisms in conditions, linking genetic variation to disease-relevant pathways. EMO integrates DNA sequence and chromatin accessibility data to predict how noncoding variants regulate gene expression across tissues and single cells, enabling context-aware personalized insights into genetic effects for precision medicine.
{"title":"Predicting the regulatory impacts of noncoding variants on gene expression through epigenomic integration across tissues and single-cell landscapes","authors":"Zhe Liu \u0000 (, ), Yihang Bao \u0000 (, ), An Gu \u0000 (, ), Weichen Song \u0000 (, ), Guan Ning Lin \u0000 (, )","doi":"10.1038/s43588-025-00878-7","DOIUrl":"10.1038/s43588-025-00878-7","url":null,"abstract":"Noncoding mutations play a critical role in regulating gene expression, yet predicting their effects across diverse tissues and cell types remains a challenge. Here we present EMO, a transformer-based model that integrates DNA sequence with chromatin accessibility data (assay for transposase-accessible chromatin with sequencing) to predict the regulatory impact of noncoding single nucleotide polymorphisms on gene expression. A key component of EMO is its ability to incorporate personalized functional genomic profiles, enabling individual-level and disease-contextual predictions and addressing critical limitations of current approaches. EMO generalizes across tissues and cell types by modeling both short- and long-range regulatory interactions and capturing dynamic gene expression changes associated with disease progression. In benchmark evaluations, the pretraining-based EMO framework outperformed existing models, with fine-tuning small-sample tissues enhancing the model’s ability to fit target tissues. In single-cell contexts, EMO accurately identified cell-type-specific regulatory patterns and successfully captured the effects of disease-associated single nucleotide polymorphisms in conditions, linking genetic variation to disease-relevant pathways. EMO integrates DNA sequence and chromatin accessibility data to predict how noncoding variants regulate gene expression across tissues and single cells, enabling context-aware personalized insights into genetic effects for precision medicine.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 10","pages":"927-939"},"PeriodicalIF":18.3,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145180816","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}
Pub Date : 2025-09-26DOI: 10.1038/s43588-025-00876-9
Olesia Dogonasheva, Keith B. Doelling, Denis Zakharov, Anne-Lise Giraud, Boris Gutkin
Unraveling how humans understand speech despite distortions has long intrigued researchers. A prominent hypothesis highlights the role of multiple endogenous brain rhythms in forming the computational context to predict speech structure and content. Yet how neural processes may implement rhythm-based context formation remains unclear. Here we propose the brain rhythm-based inference model (BRyBI) as a possible neural implementation of speech processing in the auditory cortex based on the interaction of endogenous brain rhythms in a predictive coding framework. BRyBI encodes key rhythmic processes for parsing spectro-temporal representations of the speech signal into phoneme sequences and to govern the formation of the phrasal context. BRyBI matches patterns of human performance in speech recognition tasks and explains contradictory experimental observations of rhythms during speech listening and their dependence on the informational aspect of speech (uncertainty and surprise). This work highlights the computational role of multiscale brain rhythms in predictive speech processing. This study presents a brain rhythm-based inference model (BRyBI) for speech processing in the auditory cortex. BRyBI shows how rhythmic neural activity enables robust speech processing by dynamically predicting context and elucidates mechanistic principles that allow robust speech parsing in the brain.
{"title":"Rhythm-based hierarchical predictive computations support acoustic−semantic transformation in speech processing","authors":"Olesia Dogonasheva, Keith B. Doelling, Denis Zakharov, Anne-Lise Giraud, Boris Gutkin","doi":"10.1038/s43588-025-00876-9","DOIUrl":"10.1038/s43588-025-00876-9","url":null,"abstract":"Unraveling how humans understand speech despite distortions has long intrigued researchers. A prominent hypothesis highlights the role of multiple endogenous brain rhythms in forming the computational context to predict speech structure and content. Yet how neural processes may implement rhythm-based context formation remains unclear. Here we propose the brain rhythm-based inference model (BRyBI) as a possible neural implementation of speech processing in the auditory cortex based on the interaction of endogenous brain rhythms in a predictive coding framework. BRyBI encodes key rhythmic processes for parsing spectro-temporal representations of the speech signal into phoneme sequences and to govern the formation of the phrasal context. BRyBI matches patterns of human performance in speech recognition tasks and explains contradictory experimental observations of rhythms during speech listening and their dependence on the informational aspect of speech (uncertainty and surprise). This work highlights the computational role of multiscale brain rhythms in predictive speech processing. This study presents a brain rhythm-based inference model (BRyBI) for speech processing in the auditory cortex. BRyBI shows how rhythmic neural activity enables robust speech processing by dynamically predicting context and elucidates mechanistic principles that allow robust speech parsing in the brain.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 10","pages":"915-926"},"PeriodicalIF":18.3,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145180835","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}
Pub Date : 2025-09-24DOI: 10.1038/s43588-025-00890-x
This issue of Nature Computational Science features a Focus that highlights both the promises and perils of large language models, their emerging applications across diverse scientific domains, and the opportunities to overcome the challenges that lie ahead.
{"title":"The rise of large language models","authors":"","doi":"10.1038/s43588-025-00890-x","DOIUrl":"10.1038/s43588-025-00890-x","url":null,"abstract":"This issue of Nature Computational Science features a Focus that highlights both the promises and perils of large language models, their emerging applications across diverse scientific domains, and the opportunities to overcome the challenges that lie ahead.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 9","pages":"689-690"},"PeriodicalIF":18.3,"publicationDate":"2025-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.comhttps://www.nature.com/articles/s43588-025-00890-x.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145129541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-16DOI: 10.1038/s43588-025-00868-9
Nathan Leroux, Jan Finkbeiner, Emre Neftci
Strong barriers remain between neuromorphic engineering and machine learning, especially with regard to recent large language models (LLMs) and transformers. This Comment makes the case that neuromorphic engineering may hold the keys to more efficient inference with transformer-like models.
{"title":"Neuromorphic principles in self-attention hardware for efficient transformers","authors":"Nathan Leroux, Jan Finkbeiner, Emre Neftci","doi":"10.1038/s43588-025-00868-9","DOIUrl":"10.1038/s43588-025-00868-9","url":null,"abstract":"Strong barriers remain between neuromorphic engineering and machine learning, especially with regard to recent large language models (LLMs) and transformers. This Comment makes the case that neuromorphic engineering may hold the keys to more efficient inference with transformer-like models.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 9","pages":"708-710"},"PeriodicalIF":18.3,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145076704","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}
Pub Date : 2025-09-16DOI: 10.1038/s43588-025-00861-2
Eva Portelance, Masoud Jasbi
Chomsky’s generative linguistics has made substantial contributions to cognitive science and symbolic artificial intelligence. With the rise of neural language models, however, the compatibility between generative artificial intelligence and generative linguistics has come under debate. Here we outline three ways in which generative artificial intelligence aligns with and supports the core ideas of generative linguistics. In turn, generative linguistics can provide criteria to evaluate and improve neural language models as models of human language and cognition. This Perspective discusses that generative AI aligns with generative linguistics by showing that neural language models (NLMs) are formal generative models. Furthermore, generative linguistics offers a framework for evaluating and improving NLMs.
{"title":"On the compatibility of generative AI and generative linguistics","authors":"Eva Portelance, Masoud Jasbi","doi":"10.1038/s43588-025-00861-2","DOIUrl":"10.1038/s43588-025-00861-2","url":null,"abstract":"Chomsky’s generative linguistics has made substantial contributions to cognitive science and symbolic artificial intelligence. With the rise of neural language models, however, the compatibility between generative artificial intelligence and generative linguistics has come under debate. Here we outline three ways in which generative artificial intelligence aligns with and supports the core ideas of generative linguistics. In turn, generative linguistics can provide criteria to evaluate and improve neural language models as models of human language and cognition. This Perspective discusses that generative AI aligns with generative linguistics by showing that neural language models (NLMs) are formal generative models. Furthermore, generative linguistics offers a framework for evaluating and improving NLMs.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 9","pages":"745-753"},"PeriodicalIF":18.3,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145076692","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}
Transformer-based large language models (LLMs) have considerably advanced our understanding of how meaning is represented in the human brain; however, the validity of increasingly large LLMs is being questioned due to their extensive training data and their ability to access context thousands of words long. In this study we investigated whether instruction tuning—another core technique in recent LLMs that goes beyond mere scaling—can enhance models’ ability to capture linguistic information in the human brain. We compared base and instruction-tuned LLMs of varying sizes against human behavioral and brain activity measured with eye-tracking and functional magnetic resonance imaging during naturalistic reading. We show that simply making LLMs larger leads to a closer match with the human brain than fine-tuning them with instructions. These finding have substantial implications for understanding the cognitive plausibility of LLMs and their role in studying naturalistic language comprehension. Larger LLMs’ self-attention more accurately predicts readers’ regressive saccades and fMRI responses in language regions, whereas instruction tuning adds no benefit.
{"title":"Increasing alignment of large language models with language processing in the human brain","authors":"Changjiang Gao, Zhengwu Ma, Jiajun Chen, Ping Li, Shujian Huang, Jixing Li","doi":"10.1038/s43588-025-00863-0","DOIUrl":"10.1038/s43588-025-00863-0","url":null,"abstract":"Transformer-based large language models (LLMs) have considerably advanced our understanding of how meaning is represented in the human brain; however, the validity of increasingly large LLMs is being questioned due to their extensive training data and their ability to access context thousands of words long. In this study we investigated whether instruction tuning—another core technique in recent LLMs that goes beyond mere scaling—can enhance models’ ability to capture linguistic information in the human brain. We compared base and instruction-tuned LLMs of varying sizes against human behavioral and brain activity measured with eye-tracking and functional magnetic resonance imaging during naturalistic reading. We show that simply making LLMs larger leads to a closer match with the human brain than fine-tuning them with instructions. These finding have substantial implications for understanding the cognitive plausibility of LLMs and their role in studying naturalistic language comprehension. Larger LLMs’ self-attention more accurately predicts readers’ regressive saccades and fMRI responses in language regions, whereas instruction tuning adds no benefit.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 11","pages":"1080-1090"},"PeriodicalIF":18.3,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.comhttps://www.nature.com/articles/s43588-025-00863-0.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145076701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-12DOI: 10.1038/s43588-025-00864-z
Ying Li, Yuzhuo Ma, He Xu, Yaoyao Sun, Min Zhu, Weihua Yue, Wei Zhou, Wenjian Bi
With the growing availability of time-stamped electronic health records linked to genetic data in large biobanks and cohorts, time-to-event phenotypes are increasingly studied in genome-wide association studies. Although numerous Cox-regression-based methods have been proposed for a large-scale genome-wide association study, case ascertainment in time-to-event phenotypes has not been well addressed. Here we propose a computationally efficient Cox-based method, named WtCoxG, that accounts for case ascertainment by fitting a weighted Cox proportional hazards null model. A hybrid strategy incorporating saddlepoint approximation largely increases its accuracy when analyzing low-frequency and rare variants. Notably, by leveraging external minor allele frequencies from public resources, WtCoxG further boosts statistical power. Extensive simulation studies demonstrated that WtCoxG is more powerful than ADuLT and other Cox-based methods, while effectively controlling type I error rates. UK Biobank real data analysis validated that leveraging external minor allele frequencies contributes to the power gains of WtCoxG compared with ADuLT in the analysis of type 2 diabetes and coronary atherosclerosis. This study introduces WtCoxG, an efficient genetic analysis method for time-to-event data, which improves statistical power by addressing case ascertainment and leveraging external allele frequency information.
{"title":"Applying weighted Cox regression to genome-wide association studies of time-to-event phenotypes","authors":"Ying Li, Yuzhuo Ma, He Xu, Yaoyao Sun, Min Zhu, Weihua Yue, Wei Zhou, Wenjian Bi","doi":"10.1038/s43588-025-00864-z","DOIUrl":"10.1038/s43588-025-00864-z","url":null,"abstract":"With the growing availability of time-stamped electronic health records linked to genetic data in large biobanks and cohorts, time-to-event phenotypes are increasingly studied in genome-wide association studies. Although numerous Cox-regression-based methods have been proposed for a large-scale genome-wide association study, case ascertainment in time-to-event phenotypes has not been well addressed. Here we propose a computationally efficient Cox-based method, named WtCoxG, that accounts for case ascertainment by fitting a weighted Cox proportional hazards null model. A hybrid strategy incorporating saddlepoint approximation largely increases its accuracy when analyzing low-frequency and rare variants. Notably, by leveraging external minor allele frequencies from public resources, WtCoxG further boosts statistical power. Extensive simulation studies demonstrated that WtCoxG is more powerful than ADuLT and other Cox-based methods, while effectively controlling type I error rates. UK Biobank real data analysis validated that leveraging external minor allele frequencies contributes to the power gains of WtCoxG compared with ADuLT in the analysis of type 2 diabetes and coronary atherosclerosis. This study introduces WtCoxG, an efficient genetic analysis method for time-to-event data, which improves statistical power by addressing case ascertainment and leveraging external allele frequency information.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 11","pages":"1064-1079"},"PeriodicalIF":18.3,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145056460","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}
Pub Date : 2025-09-12DOI: 10.1038/s43588-025-00867-w
Peiyi He, Shengbo Wang, Ruibin Mao, Mingrui Jiang, Sebastian Siegel, Giacomo Pedretti, Jim Ignowski, John Paul Strachan, Ruibang Luo, Can Li
Advances in third-generation sequencing have enabled portable and real-time genomic sequencing, but real-time data processing remains a bottleneck, hampering on-site genomic analysis. These technologies generate noisy analog signals that traditionally require basecalling and read mapping, both demanding costly data movement on von Neumann hardware. Here, to overcome this, we present a memristor-based hardware–software codesign that processes raw sequencer signals directly in analog memory, combining the two separated steps. By exploiting intrinsic device noise for locality-sensitive hashing and implementing parallel approximate searches in content-addressable memory, we experimentally showcase on-site applications, including infectious disease detection and metagenomic classification on a fully integrated memristor chip. Our experimentally validated analysis confirms the effectiveness of this approach on real-world tasks, achieving a 97.15% F1 score in virus raw signal mapping, with 51× speed-up and 477× energy saving over an application-specific integrated circuit. These results demonstrate that in-memory computing hardware provides a viable solution for integration with portable sequencers, enabling real-time and on-site genomic analysis. The authors report a memristor-based system that analyzes raw analog signals from a genomic sequencer directly in memory. By bypassing slow data conversion, the system achieves substantial improvements in speed and efficiency, enabling real-time, on-site genomic analysis.
{"title":"Real-time raw signal genomic analysis using fully integrated memristor hardware","authors":"Peiyi He, Shengbo Wang, Ruibin Mao, Mingrui Jiang, Sebastian Siegel, Giacomo Pedretti, Jim Ignowski, John Paul Strachan, Ruibang Luo, Can Li","doi":"10.1038/s43588-025-00867-w","DOIUrl":"10.1038/s43588-025-00867-w","url":null,"abstract":"Advances in third-generation sequencing have enabled portable and real-time genomic sequencing, but real-time data processing remains a bottleneck, hampering on-site genomic analysis. These technologies generate noisy analog signals that traditionally require basecalling and read mapping, both demanding costly data movement on von Neumann hardware. Here, to overcome this, we present a memristor-based hardware–software codesign that processes raw sequencer signals directly in analog memory, combining the two separated steps. By exploiting intrinsic device noise for locality-sensitive hashing and implementing parallel approximate searches in content-addressable memory, we experimentally showcase on-site applications, including infectious disease detection and metagenomic classification on a fully integrated memristor chip. Our experimentally validated analysis confirms the effectiveness of this approach on real-world tasks, achieving a 97.15% F1 score in virus raw signal mapping, with 51× speed-up and 477× energy saving over an application-specific integrated circuit. These results demonstrate that in-memory computing hardware provides a viable solution for integration with portable sequencers, enabling real-time and on-site genomic analysis. The authors report a memristor-based system that analyzes raw analog signals from a genomic sequencer directly in memory. By bypassing slow data conversion, the system achieves substantial improvements in speed and efficiency, enabling real-time, on-site genomic analysis.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 10","pages":"940-951"},"PeriodicalIF":18.3,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145056487","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}
Pub Date : 2025-09-12DOI: 10.1038/s43588-025-00866-x
Tao Yan, Yanchen Guo, Tiankuang Zhou, Guocheng Shao, Shanglong Li, Ruqi Huang, Qionghai Dai, Lu Fang
The field of photonic neural networks has experienced substantial growth, driven by its potential to enable ultrafast artificial intelligence inference and address the escalating demand for computing speed and energy efficiency. However, realizing nonlinearity-complete all-optical neurons is still challenging, constraining the performance of photonic neural networks. Here we report a complete photonic integrated neuron (PIN) with spatiotemporal feature learning capabilities and reconfigurable structures for nonlinear all-optical computing. By interleaving the spatiotemporal dimension of photons and leveraging the Kerr effect, PIN performs high-order temporal convolution and all-optical nonlinear activation monolithically on a silicon-nitride photonic chip, achieving neuron completeness of weighted interconnects and nonlinearities. We develop the PIN chip system and demonstrate its remarkable performance in high-accuracy image classification and human motion generation. PIN enables ultrafast spatialtemporal processing with a latency as low as 240 ps, paving the way for advancing machine intelligence into the subnanosecond regime. This study reports a complete photonic neuron integrated on a silicon-nitride chip, enabling ultrafast all-optical computing with nonlinear multi-kernel convolution for image recognition and motion generation.
{"title":"A complete photonic integrated neuron for nonlinear all-optical computing","authors":"Tao Yan, Yanchen Guo, Tiankuang Zhou, Guocheng Shao, Shanglong Li, Ruqi Huang, Qionghai Dai, Lu Fang","doi":"10.1038/s43588-025-00866-x","DOIUrl":"10.1038/s43588-025-00866-x","url":null,"abstract":"The field of photonic neural networks has experienced substantial growth, driven by its potential to enable ultrafast artificial intelligence inference and address the escalating demand for computing speed and energy efficiency. However, realizing nonlinearity-complete all-optical neurons is still challenging, constraining the performance of photonic neural networks. Here we report a complete photonic integrated neuron (PIN) with spatiotemporal feature learning capabilities and reconfigurable structures for nonlinear all-optical computing. By interleaving the spatiotemporal dimension of photons and leveraging the Kerr effect, PIN performs high-order temporal convolution and all-optical nonlinear activation monolithically on a silicon-nitride photonic chip, achieving neuron completeness of weighted interconnects and nonlinearities. We develop the PIN chip system and demonstrate its remarkable performance in high-accuracy image classification and human motion generation. PIN enables ultrafast spatialtemporal processing with a latency as low as 240 ps, paving the way for advancing machine intelligence into the subnanosecond regime. This study reports a complete photonic neuron integrated on a silicon-nitride chip, enabling ultrafast all-optical computing with nonlinear multi-kernel convolution for image recognition and motion generation.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 12","pages":"1202-1213"},"PeriodicalIF":18.3,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145056517","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}
Pub Date : 2025-09-12DOI: 10.1038/s43588-025-00856-z
Jonah Rosenblum, Juechu Dong, Satish Narayanasamy
Genomic data from a single institution lacks global diversity representation, especially for rare variants and diseases. Confidential computing can enable collaborative genome-wide association studies (GWAS) without compromising privacy or accuracy. However, due to limited secure memory space and performance overheads, previous solutions fail to support widely used regression methods. Here we present SECRET-GWAS—a rapid, privacy-preserving, population-scale, collaborative GWAS tool. We discuss several system optimizations, including streaming, batching, data parallelization and reducing trusted hardware overheads to efficiently scale linear and logistic regression to over a thousand processor cores on an Intel SGX-based cloud platform. In addition, we protect SECRET-GWAS against several hardware side-channel attacks. SECRET-GWAS is an open-source tool and works with the widely used Hail genomic analysis framework. Our experiments on Azure’s Confidential Computing platform demonstrate that SECRET-GWAS enables multivariate linear and logistic regression GWAS queries on population-scale datasets from ten independent sources in just 4.5 and 29 minutes, respectively. Secure collaborative genome-wide association studies (GWAS) with population-scale datasets address gaps in genomic data. This work proposes SECRET-GWAS and system optimizations that overcome resource constraints and exploit parallelism, while maintaining privacy and accuracy.
{"title":"Confidential computing for population-scale genome-wide association studies with SECRET-GWAS","authors":"Jonah Rosenblum, Juechu Dong, Satish Narayanasamy","doi":"10.1038/s43588-025-00856-z","DOIUrl":"10.1038/s43588-025-00856-z","url":null,"abstract":"Genomic data from a single institution lacks global diversity representation, especially for rare variants and diseases. Confidential computing can enable collaborative genome-wide association studies (GWAS) without compromising privacy or accuracy. However, due to limited secure memory space and performance overheads, previous solutions fail to support widely used regression methods. Here we present SECRET-GWAS—a rapid, privacy-preserving, population-scale, collaborative GWAS tool. We discuss several system optimizations, including streaming, batching, data parallelization and reducing trusted hardware overheads to efficiently scale linear and logistic regression to over a thousand processor cores on an Intel SGX-based cloud platform. In addition, we protect SECRET-GWAS against several hardware side-channel attacks. SECRET-GWAS is an open-source tool and works with the widely used Hail genomic analysis framework. Our experiments on Azure’s Confidential Computing platform demonstrate that SECRET-GWAS enables multivariate linear and logistic regression GWAS queries on population-scale datasets from ten independent sources in just 4.5 and 29 minutes, respectively. Secure collaborative genome-wide association studies (GWAS) with population-scale datasets address gaps in genomic data. This work proposes SECRET-GWAS and system optimizations that overcome resource constraints and exploit parallelism, while maintaining privacy and accuracy.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"5 9","pages":"825-835"},"PeriodicalIF":18.3,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145056513","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：