Latest news with #GTEx
India Today
2 days ago
- Science
- India Today
Google DeepMind unveils AlphaGenome, AI to decode how DNA changes impact human health
Recently, when Demis Hassabis, the CEO of Google DeepMind won Nobel Prize he did not win it for physics. Instead, he won in for chemistry, specifically the kind of chemistry that goes into our bodies inside our genes and DNA. Now, DeepMind has introduced a new artificial intelligence model AlphaGenome. This is a new AI model that is specifically tuned to accurately predict how individual mutations in human DNA affect their functions. advertisementIn other words, the AI will help scientists and doctors better understanding of genome functions. This advanced AI model, developed by DeepMind, is said to bring a major leap in the research around genome. The human genome is a complete set of genetic instructions. Think of it as a comprehensive instruction manual which has the data to build and operate a living thing. Its genetic material primarily consists of DNA. Genomes can influence everything from physical traits of a human being to possible risks of diseases like cancer. A small change in our DNA can bring in major effects to our health. However right now understanding of the genes and how these changes work at a molecular level is one of the biggest challenges for biologists. DeepMinds AlphaGenome AI aims to help researchers solve these answers by providing deeper insights into genome mechanisms, especially in the parts that don't directly code for proteins but still play critical roles in regulating our AlphaGenome worksadvertisement DeepMind explains that at its core, AlphaGenome is particularly unique is its ability to evaluate both common and rare genetic variants, which are the small changes in our DNA that make each person unique. This, according to the company, is made possible by major technical advances that let the model analyze extremely long DNA sequences — up to 1 million base pairs — and generate highly detailed predictions. More importantly, the AI model can do this across many different cell types and biological processes, all within a single reveals that the new AlphaGenome was trained using extensive public datasets from large consortia such as ENCODE, GTEx, 4D Nucleome, and FANTOM5 which meticulously measured these properties across numerous human and mouse cell now, researchers often relied on multiple tools to study how genetic mutations affect different aspects of gene regulation. However, according to DeepMind AlphaGenome changes this process. It combines several capabilities into one model, reducing the need for fragmented approaches and enabling faster, more comprehensive AlphaGenome builds upon Google DeepMind's earlier genomics model, Enformer, and complements AlphaMissense, which specializes in analyzing variants within the 2 per cent of the genome that codes for proteins. The company highlights that AlphaGenome offers a vital new perspective for interpreting the vast remaining 98 percent—the non-coding regions—which are essential for orchestrating gene activity and contain many variants linked to company highlights that researchers are already using AlphaGenome to explore how certain genetic mutations may lead to cancer. In one test, it accurately predicted how a mutation linked to leukaemia could activate a harmful gene, confirming previous experimental availability AlphaGenome is currently available through an API for non-commercial, research-focused use. While it is not approved for clinical diagnosis, Google says the AI tool can help scientists identify which mutations are most likely to cause disease. However the company notes that the model is still evolving, and future versions may cover more species, cell types, or biological processes. - Ends
Business Wire
13-05-2025
- Business
- Business Wire
Valinor Discovery Launches to Simulate Drug Performance in Virtual Patients
SALT LAKE CITY--(BUSINESS WIRE)--Valinor Discovery, a company pioneering machine learning-powered simulation of therapeutic performance to derisk drug development, today exited stealth to announce their mission to train the first clinically translatable models on matched primary cell and clinical assay datasets. Valinor is currently generating data from proprietary matched multi-omics and oncology-focused clinical assay datasets to create virtual patient profiles that can be used to accelerate clinical drug development. Valinor's approach represents an exciting step forward in predictive drug development. Share 'Valinor was founded on the belief that with enough patient-derived data, machine learning will transform clinical drug development from a slow, costly, and high-risk undertaking into a streamlined process where therapeutic efficacy against clinical endpoints has already been simulated computationally,' said Josh Pacini, Co-Founder and CEO of Valinor Discovery. 'By linking molecular perturbations to actual clinical assay data, we will empower drug hunters to test therapies virtually in a matter of weeks before investing years at the bench or in the clinic,' said Zhanel Nugmanova, Co-Founder and Chief Scientific Officer of Valinor Discovery. Valinor is excited to announce its collaboration with the Computational Health Center at Helmholtz Munich to develop new industry benchmarks for drug perturbation models. Reliable benchmarks for drug perturbation models across drug modalities and therapeutic areas are crucial for the biotech industry. Professor Fabian Theis and his research team are leaders in applying machine learning models to perturbation prediction and single-cell genomics. "Valinor's approach represents an exciting step forward in predictive drug development. We are thrilled to collaborate with them to develop more robust and clinically translatable benchmarks for drug perturbation prediction models, which we will add to the popular OpenProblems platform," said Dr. Theis. Valinor is also announcing a collaboration with the Montgomery Lab at Stanford Medicine, led by Dr. Stephen Montgomery, Professor of Pathology, Genetics, and Biomedical Data Science and a leading expert in functional genomics. Renowned for its contributions to the Genotype-Tissue Expression (GTEx) project and transcriptomic studies of complex disease, Dr. Montgomery's lab will work with Valinor to apply its perturbation models to identify and assess compounds with the potential to treat Alzheimer's Disease. Valinor has also partnered with Latch Bio to offer a fully compliant, hosted web portal that gives Valinor customers easy access to their model and supporting plug-and-play preclinical and clinical workflow automation tools. In addition to actively generating its own patient-matched pre- and post-treatment multi-omics datasets, Valinor is actively collaborating with leading -omics sequencing companies to provide clinical-stage biopharma with custom models trained on data from their ongoing and past clinical trials. Strategic and Scientific Advisors Appointed Valinor has also recruited a number of biopharma and academic leaders as advisors, including current and former executives in machine learning, business development, clinical operations, and drug discovery, they include: Stephen Montgomery, PhD, Head of the Montgomery Lab and Endowed Professor of Pathology, Genetics, and Biomedical Data Science, Stanford University Chase Neumann, PhD, Associate Director of Oncology, Recursion Matt Donne, PhD, Former Head of Operations and Chief of Staff, Spring Science Bryan Norman, PhD, Former SVP of Lead Generation Chemistry at Enveda Biosciences and Senior Researcher, Eli Lilly Tim Sullivan, PhD, Chief Business Officer, Infinimmune About Valinor Discovery Valinor Discovery is building a new foundation for therapeutic R&D using generative machine learning combined with matched pre- and post- treatment multi-omics data longitudinally collected from the same patients. Valinor is currently developing a suite of models to empower drug developers to predict transcriptomic shifts, protein abundance levels, methylation changes, and clinical assay outcomes before a single patient is dosed. By reducing trial-and-error and de-risking development, Valinor aims to accelerate breakthroughs across therapeutic areas, starting with oncology. To learn more about Valinor's approach or collaborate with them on customized perturbation models for specific assets, diseases, or clinical endpoints, please visit and follow Valinor on LinkedIn and X.
