Latest news with #AlphaGenome
Yahoo
14-07-2025
- Science
- Yahoo
DeepMind's AlphaGenome Uses AI to Decipher Noncoding DNA for Research, Personalized Medicine
The puzzle seems impossible: take a three-billion-letter code and predict what happens if you swap a single letter. The code we're talking about—the human genome—stores most of its instructions in genetic 'dark matter,' the 98 percent of DNA that doesn't make proteins. AlphaGenome, an artificial intelligence system just released by Google DeepMind in London, aims to show how even tiny changes in those noncoding sections affect gene expression. DeepMind's newly released technology could transform how we treat genetic diseases. Though scientists long dismissed noncoding DNA as 'junk,' we now know this so-called dark matter controls when and how genes turn on or off. AlphaGenome shows promise in predicting how mutations in these regions cause diseases—from certain cancers to rare disorders where crucial proteins never get made. By revealing these hidden control switches, AlphaGenome could help researchers design therapies that target genetic conditions, potentially aiding millions of people. But to understand the complexity of the task for which AlphaGenome was created, one must consider how the definition of a 'gene' has evolved. The term, coined in 1909 to describe invisible units of heredity (as proposed by Gregor Mendel in 1865) initially carried no molecular baggage. But by the 1940s, the 'one gene, one enzyme' idea took hold. And by the 1960s, textbooks taught that for a stretch of DNA to be properly called a gene, it had to code for a specific protein. [Sign up for Today in Science, a free daily newsletter] Over the past two decades, the definition has broadened with the discoveries of genes that code for the numerous types of RNAs that don't get translated into proteins. Today a gene is considered to be any DNA segment whose RNA or protein product performs a biological function. This conceptual shift underscores the genome's real estate map: Only about 1 to 2 percent of human DNA directly codes for proteins. But with the broader definition, roughly 40 percent is gene territory. What remains unaccounted for is significant: more than a billion units of code that can determine how and how often genes get activated. Because relevant clues lie far apart and play out through complex cycles of gene regulation, decoding them has been among biology's hardest challenges. AlphaGenome's goal is to understand how these regions affect gene expression—and how even tiny changes can tilt the entire body's balance between health and disease. To do so, the AI system uses a DNA sequence with a length of up to one million letters as input—and 'predicts thousands of molecular properties characterising its regulatory activity,' according to a statement issued by DeepMind. Already, AlphaGenome has replicated results from genetics labs. In a June 2025 preprint study (which has yet to be peer-reviewed), AlphaGenome's team described using the model to run a simulation that mirrored known DNA interactions: mutations that act like rogue light switches by cranking a gene into overdrive in a certain type of leukemia. When AlphaGenome simulated interactions on a stretch of DNA containing both the gene and the mutation, it predicted the same complex chain of events that were already observed in lab experiments. Though AlphaGenome is currently available only for noncommercial testing, responses in the scientific community have been enthusiastic so far, with both biotech start-ups and university researchers publicly expressing excitement about the system's potential to accelerate research. Limits remain. AlphaGenome struggles to capture interactions that are more than 100,000 DNA letters away, can miss some tissue-specific nuances and is not designed to predict traits from a complete personal genome. Complex diseases that depend on development or environment also lie outside its direct scope. The system does suggest wide-ranging uses, however: By tracing how minute changes ripple through gene regulation, it could pinpoint the roots of genetic disorders. It could help in the design of synthetic DNA. And above all, it could offer a faster way to chart the genome's complex regulatory circuitry.
Scientific American
14-07-2025
- Science
- Scientific American
DeepMind's AlphaGenome Aims to Decode DNA's ‘Dark Matter'
The puzzle seems impossible: take a three-billion-letter code and predict what happens if you swap a single letter. The code we're talking about—the human genome—stores most of its instructions in genetic ' dark matter,' the 98 percent of DNA that doesn't make proteins. AlphaGenome, an artificial intelligence system just released by Google DeepMind in London, aims to show how even tiny changes in those noncoding sections affect gene expression. DeepMind's newly released technology could transform how we treat genetic diseases. Though scientists long dismissed noncoding DNA as 'junk,' we now know this so-called dark matter controls when and how genes turn on or off. AlphaGenome shows promise in predicting how mutations in these regions cause diseases—from certain cancers to rare disorders where crucial proteins never get made. By revealing these hidden control switches, AlphaGenome could help researchers design therapies that target genetic conditions, potentially aiding millions of people. But to understand the complexity of the task for which AlphaGenome was created, one must consider how the definition of a 'gene' has evolved. The term, coined in 1909 to describe invisible units of heredity (as proposed by Gregor Mendel in 1865) initially carried no molecular baggage. But by the 1940s, the 'one gene, one enzyme' idea took hold. And by the 1960s, textbooks taught that for a stretch of DNA to be properly called a gene, it had to code for a specific protein. On supporting science journalism If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today. Over the past two decades, the definition has broadened with the discoveries of genes that code for the numerous types of RNAs that don't get translated into proteins. Today a gene is considered to be any DNA segment whose RNA or protein product performs a biological function. This conceptual shift underscores the genome's real estate map: Only about 1 to 2 percent of human DNA directly codes for proteins. But with the broader definition, roughly 40 percent is gene territory. What remains unaccounted for is significant: more than a billion units of code that can determine how and how often genes get activated. Because relevant clues lie far apart and play out through complex cycles of gene regulation, decoding them has been among biology's hardest challenges. AlphaGenome's goal is to understand how these regions affect gene expression—and how even tiny changes can tilt the entire body's balance between health and disease. To do so, the AI system uses a DNA sequence with a length of up to one million letters as input—and 'predicts thousands of molecular properties characterising its regulatory activity,' according to a statement issued by DeepMind. Already, AlphaGenome has replicated results from genetics labs. In a June 2025 preprint study (which has yet to be peer-reviewed), AlphaGenome's team described using the model to run a simulation that mirrored known DNA interactions: mutations that act like rogue light switches by cranking a gene into overdrive in a certain type of leukemia. When AlphaGenome simulated interactions on a stretch of DNA containing both the gene and the mutation, it predicted the same complex chain of events that were already observed in lab experiments. Though AlphaGenome is currently available only for noncommercial testing, responses in the scientific community have been enthusiastic so far, with both biotech start-ups and university researchers publicly expressing excitement about the system's potential to accelerate research. Limits remain. AlphaGenome struggles to capture interactions that are more than 100,000 DNA letters away, can miss some tissue-specific nuances and is not designed to predict traits from a complete personal genome. Complex diseases that depend on development or environment also lie outside its direct scope. The system does suggest wide-ranging uses, however: By tracing how minute changes ripple through gene regulation, it could pinpoint the roots of genetic disorders. It could help in the design of synthetic DNA. And above all, it could offer a faster way to chart the genome's complex regulatory circuitry.
Economy ME
08-07-2025
- Health
- Economy ME
AlphaGenome: How will Google DeepMind's AI model transform our understanding of the human genome?
Google DeepMind unveiled AlphaGenome, a groundbreaking artificial intelligence (AI) model poised to transform our understanding of the human genome and its impact on health, disease, and biotechnology. By leveraging state-of-the-art neural architectures and vast public genomic datasets, AlphaGenome delivers unprecedented insight into how genetic variants—both common and rare—affect gene regulation across the entire genome, not just the well-studied protein-coding regions that make up a mere 2 percent of our DNA. What is AlphaGenome? AlphaGenome is an advanced AI model developed by Google DeepMind, designed to predict how genetic variants impact gene regulation and other molecular processes at base-pair resolution across the entire genome. Unlike previous models that focused primarily on protein-coding DNA, AlphaGenome analyzes both coding and non-coding regions, offering a unified framework for interpreting the regulatory landscape of human genetics. Key highlights: Processes up to 1 million base pairs of DNA at once. Predicts thousands of molecular modalities, including gene expression, chromatin accessibility, RNA splicing, and protein binding. Integrates convolutional neural networks (CNNs) and transformers for both local motif detection and long-range genomic interactions. Trained on large-scale, multi-omic datasets (ENCODE, GTEx, 4D Nucleome, FANTOM5). Available via an API for non-commercial research, with plans for broader release. The need for advanced genomic AI The complexity of the human genome The human genome is a vast instruction manual, with over 3 billion DNA letters (base pairs). While only about 2 percent of these code for proteins, the remaining 98 percent—the non-coding regions—play crucial roles in regulating gene activity, determining when and where genes are turned on or off, and influencing susceptibility to diseases. Challenges in genomic interpretation: Variant effect prediction: Small changes (variants) in DNA can have profound or negligible effects, depending on their context. Non-coding regions: Most disease-associated variants identified by genome-wide association studies (GWAS) lie outside protein-coding regions, making their functional consequences difficult to interpret. Data volume: The scale and complexity of genomic data require models that can process long sequences and integrate diverse molecular signals. AlphaGenome was developed to address these challenges, providing a comprehensive, high-resolution view of how genetic variation shapes biology. Technical architecture of AlphaGenome Unified model for sequence-to-function prediction AlphaGenome's architecture is a hybrid neural network that combines the strengths of convolutional layers and transformer modules: Convolutional Neural Networks (CNNs): Detect short, local sequence motifs—such as transcription factor binding sites—by scanning DNA for recurring patterns. Transformers: Capture long-range dependencies and interactions between distant genomic elements, essential for modeling regulatory networks that span thousands of base pairs. This design enables AlphaGenome to analyze up to 1 million base pairs in a single pass, providing base-resolution predictions across vast genomic regions. Efficient training and inference Trained on Tensor Processing Units (TPUs), AlphaGenome achieves high computational efficiency, completing full model training in just four hours—using half the compute budget of its predecessor, Enformer. The model's architecture and data pipelines are optimized for both speed and accuracy, allowing rapid hypothesis generation and variant scoring at scale. Training data and benchmark performance Multi-omic datasets AlphaGenome's predictive power is rooted in its exposure to diverse, high-quality datasets: ENCODE: Comprehensive maps of functional elements in the genome. GTEx: Gene expression data across tissues. 4D Nucleome: Insights into genome structure and organization. FANTOM5: Transcriptional activity data. Benchmarking results Outperformed or matched specialized models in 24 out of 26 benchmark tests for variant effect prediction. Demonstrated superior performance in predicting regulatory effects, RNA splicing, and chromatin accessibility. Achieved state-of-the-art results in both single-sequence and variant effect prediction tasks. Key features and innovations Comprehensive variant effect prediction AlphaGenome can score both common and rare variants across the genome, including: Non-coding regulatory regions: Where most disease-associated variants reside. Protein-coding regions: Complementing tools like AlphaMissense. Multi-modal, base-resolution output Provides predictions for thousands of molecular properties at single-base resolution, enabling fine-grained analysis of genetic changes. Models RNA splice junctions directly—a critical advance for understanding diseases caused by splicing errors. Long-range genomic context Captures interactions between distant regulatory elements, such as enhancers and promoters, which are essential for accurate gene regulation modeling. Efficient, scalable, and accessible Trained efficiently on TPUs, with rapid inference capabilities. Available via API for non-commercial research, democratizing access for scientists worldwide. Applications in genomic research Decoding the non-coding genome AlphaGenome's ability to interpret the 98 percent of the genome that does not code for proteins opens new avenues for: Identifying regulatory variants that influence gene expression and disease risk. Prioritizing candidate variants in genome-wide association studies (GWAS). Understanding tissue-specific gene regulation and its disruption in disease. Functional genomics and hypothesis generation Researchers can use AlphaGenome to: Predict the impact of specific mutations before experimental validation. Generate functional hypotheses at scale, accelerating discovery in genetics and molecular biology. Impact on disease understanding and precision medicine From variant to function to disease AlphaGenome bridges the gap between genetic variation and biological function, providing insights that are crucial for: Rare disease diagnosis: Interpreting the effects of unique or de novo variants in patients with undiagnosed conditions. Cancer genomics: Understanding how somatic mutations in regulatory regions drive tumorigenesis. Pharmacogenomics: Predicting individual responses to drugs based on regulatory variants. Toward personalized medicine By enabling accurate prediction of variant effects across tissues and cell types, AlphaGenome supports the development of personalized therapies and precision diagnostics tailored to each individual's unique genetic makeup. Read more: UAE healthcare sector aims for 20 percent carbon emission reduction by 2030: Report Synthetic biology and beyond Designing synthetic DNA AlphaGenome's predictive capabilities extend to synthetic biology, where researchers aim to design custom DNA sequences with desired regulatory properties: Synthetic promoters and enhancers: Engineering regulatory elements for gene therapy or industrial biotechnology. Genome editing: Anticipating the consequences of CRISPR and other genome-editing interventions. Expanding to other species DeepMind has indicated plans to extend AlphaGenome's framework to new species , facilitating comparative genomics and cross-species functional annotation. AlphaGenome vs. previous models Feature AlphaGenome Enformer (2022) AlphaMissense (2023) Sequence length Up to 1 million bp Up to 200,000 bp N/A (missense focus) Coding & non-coding regions Yes Yes Coding only Variant effect prediction Yes (all regions) Limited Missense only Multi-modal output Thousands of types Dozens Protein function Splice junction modeling Direct Indirect No Training efficiency 4 hours on TPUs 8+ hours N/A Benchmark performance 24/26 top scores 18/26 N/A AlphaGenome represents a substantial leap in both scale and accuracy compared to previous models, especially in non-coding variant interpretation and multi-modal prediction. Ethical, societal, and clinical considerations Interpretability and trust As AI models become central to genomic interpretation, issues of transparency, explainability, and clinical validation are paramount. AlphaGenome's predictions must be interpreted within the context of experimental evidence and patient care, with careful attention to: False positives/negatives in variant effect prediction. Equity and access to advanced genomic tools across different populations and healthcare systems. Data privacy and security Handling genomic data raises significant privacy concerns, necessitating robust safeguards for patient information and compliance with global regulations. The human element As noted by AI alignment researchers, the psychological and informational context in which genomic insights are delivered is as important as their technical accuracy. AI must support clinicians in providing clear, compassionate communication to patients. The road ahead: Future developments Clinical integration DeepMind plans to extend AlphaGenome for clinical applications, including fine-tuning for disease-specific tasks, integration with electronic health records, and support for clinical decision-making. Expansion to other organisms and modalities Ongoing work aims to adapt AlphaGenome for other species and new molecular phenotypes, broadening its impact across biology and medicine. Open science and collaboration By making AlphaGenome available via API for non-commercial research, DeepMind promotes global collaboration and accelerates discovery in genomics. Final word AlphaGenome marks a new era in computational genomics, offering a unified, scalable, and accurate framework for interpreting the functional consequences of genetic variation across the entire genome. Its release in 2025 represents a milestone not just for AI and genomics, but for the broader quest to understand the language of life and harness it for human health, disease prevention, and biotechnological innovation.
Sustainability Times
29-06-2025
- Science
- Sustainability Times
'Google Just Changed Everything': This Ruthless New AI Reads 1 Million Human DNA Letters Instantly and Scientists Are Stunned
IN A NUTSHELL 🔬 Google unveils the AlphaGenome model, revolutionizing genomics with its ability to analyze one million DNA letters in a single pass. model, revolutionizing genomics with its ability to analyze one million DNA letters in a single pass. 💡 The model uses a combination of convolutional layers and transformers to predict molecular properties like gene expression and splicing patterns. and to predict molecular properties like gene expression and splicing patterns. 🧬 AlphaGenome advances disease research by accurately predicting the impact of genetic variations, aiding in the study of rare variants and RNA splicing disruptions. advances disease research by accurately predicting the impact of genetic variations, aiding in the study of rare variants and RNA splicing disruptions. 🚀 With potential applications in synthetic biology, the model helps design regulatory sequences, opening new possibilities for targeted therapies and genetic engineering. In a groundbreaking advancement for the field of genomics, Google has unveiled its latest AI innovation: the AlphaGenome model. This powerful tool marks a significant leap forward in our ability to understand genetic information, as it can analyze and predict the effects of up to one million DNA letters in a single pass. By providing base-level insights into the complex machinery that regulates gene activity, AlphaGenome offers an unprecedented view into both coding and non-coding regions of the genome. This development has the potential to unlock new insights in disease biology, synthetic DNA design, and much more. Understanding the AlphaGenome Model The AlphaGenome model employs a sophisticated architecture that combines convolutional layers and transformers to analyze DNA sequences. Initially, convolutional layers detect short patterns within the genetic code, while transformers ensure that information is shared across the entire DNA sequence. This intricate approach allows the model to predict a wide array of molecular properties, such as gene expression, splicing patterns, and protein-binding sites. A notable feature of AlphaGenome is its ability to provide base-resolution insight into long-range genomic interactions, decoding the impact of mutations at an unprecedented scale. During its training phase, AlphaGenome was fine-tuned using massive public datasets, including ENCODE, GTEx, 4D Nucleome, and FANTOM5. This extensive training enables the model to outperform or match specialized models in 24 of 26 benchmark tests. Additionally, the model's computations are distributed across multiple Tensor Processing Units (TPUs), allowing for efficient large-scale processing. Remarkably, AlphaGenome was trained in just four hours, using half the compute budget required for its predecessor, Enformer. 'Time Was Here First': Mind-Blowing Discovery Reveals the Universe Was Born from Time Itself, Not from Space at All Applications in Disease Research One of the most significant applications of AlphaGenome lies in its potential to advance disease research. The model's variant scoring system efficiently contrasts mutated and unmutated DNA, providing insights into the impact of genetic variations across different modalities. This capability is particularly valuable for studying rare variants with large biological effects, such as those responsible for Mendelian disorders. Furthermore, AlphaGenome's splice-junction modeling offers a novel approach to predicting RNA splicing disruptions, which are linked to diseases like cystic fibrosis and spinal muscular atrophy. In one test case, the model accurately predicted how a leukemia-linked mutation introduces a MYB DNA binding motif, activating the TAL1 gene. This finding mirrors known mechanisms in T-cell acute lymphoblastic leukemia, demonstrating AlphaGenome's power to connect non-coding variants to disease genes. 'Robot Did the Surgery': U.S. Doctors Complete First-Ever Heart Transplant Without Opening the Chest Implications for Synthetic Biology Beyond disease research, AlphaGenome holds promise in the field of synthetic biology. The model's ability to predict gene expression and regulatory patterns makes it a valuable tool for designing regulatory sequences that activate genes selectively. For instance, AlphaGenome could help develop sequences that activate genes in nerve cells but not in muscle cells, offering new possibilities for targeted therapies and genetic engineering. By capturing long-range genomic context and offering base-level predictions, AlphaGenome addresses a significant limitation of earlier models, which often traded sequence length for resolution. This comprehensive approach unlocks new insights into the intricate workings of the genome, paving the way for advancements in various areas of biological research. 'We Found Alien-Like Creatures Feeding on Methane': Sea Spiders Survive on Microbes in One of Earth's Strangest Ecosystems Future Prospects and Challenges While AlphaGenome represents a major advancement, it is not without its limitations. The model is not designed or validated for personal genome interpretation or clinical use, and it faces challenges in modeling very distant regulatory interactions. These limitations highlight areas for future development and refinement. Nonetheless, AlphaGenome lays a strong foundation for future expansion, with the potential to be adapted for additional species, modalities, and lab-specific datasets. Google has made the model available in preview for non-commercial use via the AlphaGenome API, inviting researchers worldwide to explore its capabilities and provide feedback. As Google states, 'We hope AlphaGenome will help deepen our understanding of the complex cellular processes encoded in DNA and drive new discoveries in genomics and healthcare.' As we look to the future, the implications of the AlphaGenome model are vast and varied. From advancing our understanding of genetic diseases to enabling new breakthroughs in synthetic biology, this powerful AI tool is poised to transform the field of genomics. What new discoveries will arise from this technological leap, and how will they shape the future of healthcare and biological research? Our author used artificial intelligence to enhance this article. Did you like it? 4.6/5 (20)
Yahoo
27-06-2025
- Business
- Yahoo
Alphabet Google's (GOOGL) DeepMind Unveils AlphaGenome to Revolutionize DNA Research
Alphabet Inc. (NASDAQ:GOOGL) is one of the . On June 25, Google's DeepMind announced the launch of AlphaGenome: an AI model that helps scientists better understand DNA, the hereditary material in all organisms. The new artificial intelligence tool by Google is designed to predict how genetic variations in human DNA impact biological processes that regulate genes. This AI model is capable of analyzing up to 1 million DNA base pairs and predicting thousands of molecular properties related to regulatory activity. It can also assess the effects of genetic variants by comparing predictions between mutated and unmutated sequences. The company is making AlphaGenome available in preview via their AlphaGenome API for non-commercial research, and also plans to release the model in the future. The predictive capabilities offered by AlphaGenome will help more accurately predict genetic disruptions, guide the design of synthetic DNA with specific regulatory function, and also accelerate our understanding of the genome. Photo by Kai Wenzel on Unsplash 'AlphaGenome will be a powerful tool for the field. Determining the relevance of different non-coding variants can be extremely challenging, particularly to do at scale. This tool will provide a crucial piece of the puzzle, allowing us to make better connections to understand diseases like cancer.' -Professor Marc Mansour, University College London Alphabet Inc. (NASDAQ:GOOG) is an American multinational technology conglomerate holding company wholly owning the internet giant Google, amongst other businesses. While we acknowledge the potential of GOOG as an investment, we believe certain AI stocks offer greater upside potential and carry less downside risk. If you're looking for an extremely undervalued AI stock that also stands to benefit significantly from Trump-era tariffs and the onshoring trend, see our free report on the best short-term AI stock. READ NEXT: 10 AI Stocks in the Spotlight and . Disclosure: None. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
